Cell Architecture Explained

IdiotOnMyLeft writes "OSNews features an article written by Nicholas Blachford about the new processor developed by IBM and Sony for their Playstation 3 console. The article goes deep inside the Cell architecture and describes why it is a revolutionary step forwards in technology and until now, the most serious threat to x86. '5 dual core Opterons directly connected via HyperTransport should be able to achieve a similar level of performance in stream processing - as a single Cell. The PlayStation 3 is expected to have have 4 Cells.'"

Comments

Seeing is believing

[Anonymous Coward]

It's not like we haven't heard it before. It usually turns out to be halfish-truish for some restricted subset of operations in a theoretical setting, you know where you discount busses, memory and latencies.

Re:Seeing is believing

[Anonymous Coward]

RTFA. He covers that in explicit detail.

Re:Seeing is believing

[Anonymous Coward]

His coverage is speculation through reading the patent(s). RTFA won't help you, if the real thing doesn't perform as expected.

For example, he keeps saying Cells are going to be so cheap it will wipe out PCs, but you think Intel and AMD aren't going to lower their prices further if nobody buys their CPUs? Some people would rather buy an expensive, lower performance, and DRM-free CPU. He also said Cell's introduction with a huge performance gap won't ever be repeated...

Re:Seeing is believing

[aphor]

No, this sort of architecture is a general trend towards paralellization. It is smart, and it is known to work, and I would expect some bright Sparc wise people to chime in and say "u-huh" and some SGI wise people to chime in and say "I've seen some of this before." The OS people are starting to move things in this direction, and I've heard that Darwin has had the asynchronous messaging type threading model for a while (RTFA: the article explicitly mentions Tiger's GPU leveraging techniques). If you have the head for it, try reading up on NUMA and compare that with SMP.

The math is simple. CPUs are CPUs, and anyone can make one that is the same speed as the competition, and if they do it second they can do it cheaper. The guy that can make 20 CPUs work like one CPU that does 20 times the work in a given time will win because he can always just throw more hardware at the problem. The SMP guys have to go back to the drawing board. In this case, the only way to beat-em is to join-em. Maybe doing the specific "Cell" computing design isn't it, but the ol' PC is dead. If these things start hitting the commodity price-points.

That's a big, fat IF. So, don't bet on it (yet), but it's even worse to ignore it.

Re:Seeing is believing

[aichpvee]

That's right, it's time for another ride on the Sony hype machine! Watch out if you live in China, though. The US government is going to make this one illegal to export there due to its status as a "supercomputer".

Re:Seeing is believing

[Raunch]

Then perhaps some of those bright people can shine some light in my direction.

FTFA
Caches work by storing part of the memory the processor is working on, if you are working on a 1MB piece of data it is likely only a small fraction of this (perhaps a few hundred bytes) will be present in cache, there are kinds of cache design which can store more or even all the data but these are not used as they are too expensive or too slow.

APU local memory - no cache
To solve the complexity associated with cache design and to increase performance the Cell designers took the radical approach of not including any. Instead they used a series of local memories, there are 8 of these, 1 in each APU.

The APUs operate on registers which are read from or written to the local memory. This local memory can access main memory in blocks of 1024 bits but the APUs cannot act directly on main memory.

By not using a caching mechanism the designers have removed the need for a lot of the complexity which goes along with a cache.

This may sound like an inflexible system which will be complex to program and it most likely is but this system will deliver data to the APU registers at a phenomenal rate. If 2 registers can be moved per cycle to or from the local memory it will in it's first incarnation deliver 147 Gigabytes per second. That's for a single APU, the aggregate bandwidth for all local memories will be over a Terabyte per second - no CPU in the consumer market has a cache which will even get close to that figure. The APUs need to be fed with data and by using a local memory based design the Cell designers have provided plenty of it.

Ok, so regular memory is too slow, or too expensive (apparently caching is the main problem although what I remember from comp. arch. was that caching was a Good Thing). So then they abolish the cache.
Caching: CPU > cache (1 or 2) > main memory > HD
and then implement another system that is *completely different*
Non-Caching: APU > local memory > main memory
Now first off, how is it different? Secondly how does this improve the physical memory speed? Is the author claiming that a page fault is what we are avoiding here? If that's the case, then the problem is hard drives not solid state memory. But (again if what I learned in comp arch is correct), because there is a tendency for programs to run the same code over and over again, then (assuming you have a good algorithm) the time saved by caching is signifigant.

Anyone?

Re:Seeing is believing

[Henk Poley]

So then they abolish the cache.
Caching: CPU > cache (1 or 2) > main memory > HD
and then implement another system that is *completely different*
Non-Caching: APU > local memory > main memory
Now first off, how is it different?

Let me take a vaguely educated guess.

Currently the cache managers in x86 CPUs "predict" what part of the memory space is needed. This prediction isn't always that good, and efforts to make programs hint to processor what to cache haven't worked good enough (or at least according to Cell CPU designers). So they force to program to operate in a 'small' memory space where data can be read in from a large RAM storage.

I don't know if it will actually help. To me it seems a bit like going back tot he 80268/80386 era with himem.drv under DOS so programs could acces higher memory regions by commanding the driver to swap in and out memory to the lower 640k.

But then, maybe Bill Gates -or whoever put up that quote in his name- was right, 640k RAM is enough for everyone (multiplied by the number of cores on dye..).

Re:Seeing is believing

[Mycroft_VIII]

The difference here is instead of cache, you have localized memory. Cache just maps some external ram adress so they pull from a small on board bank of very fast ram. This adds the complexity of making shure that the cache has the same data as the ram.
With what the cell apears to be doing you just load the whole chunk of code to be run by the an apu into its local memory. Then you don't have to worry about the slower mainboard ram untill you have something to put back, which I assume can go anywhere that's free at the moment.
Also with conventional cache schemes you have to pay attention to how the code executes to make shure ALL of that fast loop will fit in cache, and that it winds up there. Modern compilers and such do a lot (practically all if given a specific processor to target) of this work for you, but with cells what winds up in the local memory each apu has is determinable in a more direct fashion. In fact this is part of how you program them, rather than being a programming mechanism to take advantage of a crutch designed to deal with the disparities of speed between processors and normal ram.
Think of it like this, what if your pc's ram was all at full core speed and your hard-drive was as fast as 400mhz ddr. It's somewhat analogous as the apu treats the high-spead local memory like your cpu treats ram and treats the off chip ram like a hdd, it loads each task from ram into it's local memory, executes it, saves the results back out ram, or ships them over to another apu on the chip for another process to use.
At least that is kinda how I see it.

Mycroft

Re:Seeing is believing

[be-fan]

The local storage is different for the following reasons:

1) It must be programmed differently. Instead of just accessing memory how you want, you must explicitly copy the part of memory you need at the moment. So, if your APU is acting as a vertex shader, you need to copy the shader code into the LS before you start processing. Essentially, the LS can give you the time savings of a cache, but you have to manage it yourself to get the benefits.

2) Since the LS isn't managed by the hardware, it doesn't need a lot of management hardware. You don't need cache tags, lookup hardware, hardware to manage misses, etc. This saves a lot of transistors.

3) A regular cache has to do some management on each access. It has to search the tags to find what cache line holds a given memory word, it has to perform write-back, etc. Since the LS doesn't need to do any of this, latency can be cut down.

4) Since the LS is addressed directly, and isn't mapped onto memory, there is no need for cache coherency protocols. A cache-coherent multi-processor system needs to communicate with it's peers to coordinate access to the cache. For example, when it writes to a memory location, it must notify all other processors caching that location that their copies are now invalid. The LS doesn't need to do any of this, and that cuts down on both management hardware and latency.

The APUs are stream processors. It is common for stream processors to not have a general memory cache. The geforce 3's vertex processor, for example, has enough cache to hold 18 vertices. The are not an LRU cache like a Pentium's, but a FIFO (much cheaper to manage), and are only used in certain circumstances. In comparison, the Cell's 128KB LS per PE is enormous!

Re:Seeing is believing

[Henk Poley]

Since the LS is addressed directly, and isn't mapped onto memory, there is no need for cache coherency protocols. A cache-coherent multi-processor system needs to communicate with it's peers to coordinate access to the cache.

So they basicly built a "NUMA on a chip"?

Human skin cells

[MarkRose]

Yeah, but can my inkjet print them?

Re:Human skin cells

[SpaceLifeForm]

Please mods, if you don't get it, don't moderate it.

See this story.

Re:Human skin cells

[Anonymous Coward]

At least someone got it. I guess I should have linked the story ;)

PS3 As A Server?

[Anonymous Coward]

I dont know about you but I'm looking forward to putting linux on one of these.

Re:PS3 As A Server?

[Anonymous Coward]

The PS3 will be using some form of Linux for its operating system.

Re:PS3 As A Server?

[Anonymous Coward]

With PS3,
Can I compress MPEG2 to H.264 AVC Nero Digital (Max. Ratio) a 3 hours porn film in 5 minutes?

WOW!!!, the japanese men are authentic superengineers!!! (;

open4free ©

This is beginning to sound more like

[gotr00t]

a DBZ reference: "Part 4: Cell Vs the PC"

Re:This is beginning to sound more like

[goodbadorugly]

a DBZ reference: "Part 4: Cell Vs the PC"

The 45 episode saga in which:

Bill Gates becomes a cyborg and summons the forces of evil.

A new Cell is constructed out of unsold Itaniums (Not to be confused with the Cell built by Sony, which is a friendly robot that is found out to be good. ( Until he is found out to be evil when the heroes notice he is under the control of the cyborg Bill Gates who has been behind the charade the entire time) and challenges the world to a rematch of earth shattering proportions

Second string characters have meaningless conversations that take up entire episodes

There is hilarious comic relief from common citizens in various towns as their cities crumble around them

Krillin dies

The dragon is summoned

Goku gets a haircut ...Good lord I should have my anime viewers license revoked for knowing all that crap.

Re:This is beginning to sound more like

[Anonymous Coward]

Not that curious about cell. If you have instructions on building Android 18 however ...

Re:This is beginning to sound more like

[Rethcir]

You forgot that the first 10 minutes of each episode will recap the prior episode, and the last 10 minutes of each episode will be a preview of the next episode.

Inuyasha? KAGOME!

Re:This is beginning to sound more like

[A Drake Man]

Don't forget that before you can activate the processor, you have to scream for approximately 5 minutes (not constantly, you can take quick breaths). You know you're doing it right when your hair changes color and you start to glow (small rocks lifting from the ground is a good indication as well).

Re:This is beginning to sound more like

[19thNervousBreakdown]

Don't forget the 10 minutes of vibrating eyes in the middle!

Re:This is beginning to sound more like

[LifesABeach]

Is it only me, but this 'Cell' sounds more like a 'Parallel' processor with a minor variation of functionality. I remember this stuff back in the 1970's; Only the I/O channels were Lazer diodes, and Doc. Chow was in a pissing contest with Doc. Cray. Cray won, because he was the boss.

Re:This is beginning to sound more like

[trentblase]

That's.... not.... POSSSIBLE!

So

[ikkonoishi]

How long until the first beowulf cluster of ps3s?

Also can it run linux?

Re:So

[lemnik]

I wouldn't be supprised, since PS2 runs linux. It might take some time for linux to be ported but I'm sure it will be. Currently there is a special OS under development specifically to deal with the Cell CPU.

Re:So

[Anonymous Coward]

Well if you take into account the fact that it is based on power platform.

And that Sony has previous Linux experiance thru porting it to Playstation2.

And that IBM has extensive experiance at hardware virtualization.

And that Linux has been ported to dozens of other platforms and areas, especially the Power platform. (PowerPC is second favorite Linux desktop platform behind x86)

And that IBM is releasing a Cell based workstation that will probably run linux for game developers to develop on.

And IBM has a big boner for Linux, lately. (for instance they are the company that has the largest number of full time employees working specificly on Linux and free software in the world.)

That linux 2.6 has proven scalability for massive numbers of proccessors, and is the most used OS in the world for clustering.

I'd say, yes. It is very likely that Linux will run on it.

Re:So

[Anonymous Coward]

In Soviet Slashdot, "In Soviet Russia, 'In Soviet Russia' posts get modded down." posts get modded down.

Re:So

[maxwell demon]

In Soviet Russia, "In soviet Russia" posts moderate YOU down.
And of course, in Korea, only old people write "in Korea, only old people" posts.
Imagine a Beowulf cluster of Beowulf clusters!

Re:So

[h4rm0ny]

Also can it run linux?

I'm still trawling my way through the article, but tihs is a brand new, non-86 architecture. I would imagine there is a lot of work to be done on porting any O/S to work on it properly.

A more pertinant question would be can Linux run on it before Windows does? If there really were a big shift in hardware platforms, which I suppose there must be at some point, then the development speeds of different OS will really make a difference to who dominates.

If this really is a big shift in the hardware basis for modern computing, then we'll really see a fair contest between proprietary and Open Source at last.

Re:So

[Anonymous Coward]

A more pertinant question would be can Linux run on it before Windows does?

I believe so. Microsoft are renowned to take forever to release their operating systems. The only reason they released one for 64 bit systems, is due to the close relationship they have with Intel. A totally new architecture will require a total redesign of their OS. And they wouldn't want to release parts of that OS to the general public. Linux, on the other hand, has had plenty of experience porting to different architecture (plus a huge support base). So much so that most of its source code requires only one or two modifications to compile on a different architecture, and a compiler to do this. As soon as the kernel and the rest of the base system are ported, the rest will be no trouble.

If this really is a big shift in the hardware basis for modern computing, then we'll really see a fair contest between proprietary and Open Source at last.

The only problem with open source vs proprietary, is that proprietary has the money for PR. I have never seen anyone advertise any open source software on TV, although i live in South Africa, maybe it is different elsewhere.

Re:So

[Anonymous Coward]

I think linux _will_ run on the hardware (at least the G5 portion of it) with only slight modifications. The main developement will be on optimizing; writing the software cells to make things _really_ fast.

The OS will have the additional tasks of sending software cells to the individual hardware cell units for processing. It will also have to manage the input and output for those cells.

To think of a software cell lets use a unix command pipeline analogy:

gzip -cd file.tgz | tar xf -

the software cells would be software packets similar to gzip, bzip2, tar, grep, sed, gawk, perl?

The OS would mange issuing the software packets and it's data, and the redirection parts by sending the output of one cell to the input of the next in sequence?

Note that the Power series (G[45]) is oriented around memory streams that operate optimally when input and output are more balanced (the x86 is optimized for input or output bursts)

So you _can_ do serial stuff in parallel.

Someone is going to have to write a compiler or interpreter that produces cell VM code. But that propably isn't that difficult.

The part that's really cool is that there probably will only have minimal memory management, a raw memory segment to abuse, a subset of the interupts that x86 has, and NO threading problems since one hardware cell is probably going to be used exclusivly by it's software cell host until execution finishs.

Optimally there even would be a way of creating libraries for cell usage...

So the architecture seems to have all the peices in the right places. A super Mac is certainly an option; optimizing Photoshop and whatever other graphics processing would be very simple.

Re:So

[rob_osx]

What the article does not mention, or I forget if it does, is that the Cell is based on the POWER architecture. Linux and OS X are already running on the POWER architecture, and I wouldn't be suprised if they were compatible with the Cell aready. IBM wants to sell these as servers. IBM has stated that a rack of Cell servers will have 16 teraflops of performance. Yes, I believe IBM's hype, even if it is theoretical performance. In contrast a rack of Apple's Xserves only does 630 gigaflops. That's 25 times the performance. I bet IBM has Linux running on it already.

if it sounds too good to be true..

[gl4ss]

..it probably is.

was the ps2 the supercomputer it was said to be...?

the author goes on to suggest that cell workstations would smoke x86 counterparts.. but says at the same time that there probably wont be that many of them.

wtf? though in-between the lines you can read at the end that he also thinks a single g5-cpu workstation would 'smoke' x86's...

Re:if it sounds too good to be true..

[Anonymous Coward]

Looking at the ps2 now can it do toy story graphics? yes.

What it lacked when it came out was the development kits to take advantage of it.

Re:if it sounds too good to be true..

[Anonymous Coward]

"was the ps2 the supercomputer it was said to be...?"

Yes, it was.

It was just so incredibly difficult to program for, not many people got the full benefit of it :)

Re:if it sounds too good to be true..

[kai.chan]

was the ps2 the supercomputer it was said to be...?

I don't remember Sony making any big statements about the Emotion Engine being a supercomputer. What I do remember, is that when they released the clock speed of their processor, people knew the relative power of the PS2. From what I see of the Cell architecture, I can guarantee that the Cell is much more powerful than any AMD and Intel processors.

It seems like you didn't read much into the technical aspect of the Cell architecture presented in the long write-up. From just looking at a simple top-level diagram of the Cell architecture, it is clearly shown that the Cell is much more powerful than any other processors currently available. A Cell contains a Processor Unit with 8 additional Processor Units, each with its own registers. The architecture is also a distributed computing network capable of splitting tasks and computations over a wide variety of home electronics. Each Cell product you buy, you are increasing your processing power of your household. In conclusion, yes, it would smoke a x86 counterpart.

Re:if it sounds too good to be true..

[CoolGuySteve]

When it was released, the Emotion Engine in the PS2 actually would have been pretty wicked for supercomputing applications if Sony had sold a version with faster interconnects and more RAM. The processors in the PS2 are designed almost entirely to crunch vector operations, which is what most scientific codes rely on. It's really an excellent computer, it just sucks at graphics. The 4MB of uncompressed video memory and lack of hardware texture support are particularly ugly.

I suspect that the main reason there was never an Emotion Engine based cluster product was because the high performance market is tiny, especially compared to the console market, and Sony was already having trouble meeting demand with their exotic chipset when it first came out.

Anyways, I think the guy does go overboard about this new architecture. It probably will be a lot faster than PCs at certain tasks but you can only fit so many transistors in a chip. The cell stuff is cool though, it seems to fit a lot better with what most computers spend their time processing unless you're doing a lot of compiling or database operations.

Re:if it sounds too good to be true..

[Rothron the Wise]

I dug up an old article on the PS2.

The most blatant quantifiable lie is the 75 million polygon-remark. This is an highly idealized number of flat-shaded non-lighted polygons which the PS2 could possibly push if it had the cpu-power to actually supply enough meaningfull vertex data, or the memory to hold it.

My favorite part however is:
And while today's games possess limited intelligence, future games will feature characters having not only brains, but simulated senses like sight and hearing, leading to an unprecedented level of realism.

For example, the system could run interactive movies that blend Hollywood's expensive production systems with intricate story lines and characters who interact realistically with the user.

Ooooh. I gotta get me some of that!

Re:if it sounds too good to be true..

[PhotoBoy]

Don't you remember the stories about the US trying to stop Iraq and China getting their hands on PS2s because they were "super computers" and could be used to develop weapons or crack encryption? It was a very clever way of hyping the PS2 at the time.

In reality though the US definition of a super computer was out of date, and by the time the PS2 was released the top end x86 CPUs were also technically super computers.

Re:if it sounds too good to be true..

[Anonymous Coward]

"the top end x86 CPUs were also technically super computers."

Dude!

Stop hyping x86 CPUs!

Re:if it sounds too good to be true..

[Anonymous Coward]

You're right! Slay them all!

Savages...

Re:if it sounds too good to be true..

[binary42]

You just need to remember what the bottle neck was. It was memory bandwidth. In this case samrter VU programs have lead to being closer to the 75M mark (not very near but much closer than early on). The trick is in how you send geometry that is to be transformed into a display list to the GS (I use "display list" losely here). So really how the poly was shaded did impact performance but not as much as bandwidth. Texturing is still a problem but games can still look great. Look at Ratchet & Clank 3. Notice the simpler textures. Metal Gear Solid 3 also uses similar textures in almost every scene. But both push a "good" (I don't have a number) amount of geometry. Calling it a lie really is like calling any paper launch a lie. PS2 is not the only one guilty of such. It was a radical aproach to game console design that did not really pay off as well as I hope it will in the future.

Re:if it sounds too good to be true..

[Jaysyn]

Stopping China from getting them? Aren't they made in China?

Jaysyn

Re:if it sounds too good to be true..

[Anonymous Coward]

Looking at the ps2 now can it do toy story graphics? yes.

No. Not even close.

Re:if it sounds too good to be true..

[Anonymous Coward]

The keyword here is:

"...should be able to achieve a similar level of performance in stream processing - as a single Cell. "

"in stream processing" .. wtf? Well Duh. That's already the case versus your graphics card. Imagine trying to run Dooom II without a graphics card in software rendering mode and used an opteron to do the work of a graphics card.. so, even with two opterons in a box ..what sort of performance degradation will you expect in Doom III? (more than 5x I'd bet).

Yes Sony, Opterons make shitty graphics card. We already knew that.

Re:if it sounds too good to be true..

[jallen02]

Interesting that most business applications that get work done from day to day are basically nothing but database operations. Sure there is a lot of processing and such in some businesses, but it is mostly about tracking and managing data for analysis or improvement of business process.

Jeremy

Re:if it sounds too good to be true..

[gnuLNX]

"though in-between the lines you can read at the end that he also thinks a single g5-cpu workstation would 'smoke' x86's..."

and?????

Re:if it sounds too good to be true..

[Anonymous Coward]

They are now, they weren't originally though. Sony set up another factory when demand sky rocketed a few Christmases ago.

Re:if it sounds too good to be true..

[yasth]

Always doubt PR. At one thing the cell architecture should be very good. If the numbers they are turning in are right it is going to be a very powerful player on the gigabit/10gigabit switch market. A cut down version would make a great set top box. I myself am itching to get one to try on some matrix based AI algos. It also sounds like it has a L-tuple implemntation that should allow for some really fun distribution schemes.

Yet all this existed in some form already. You can buy massively powerful DSP cards to plug into your computer and gain some real powerful compute ability. The problem is that not all problems are easily "streamable", and even problems that are streamable, are not often cost free in the conversion. Not only that, but it is a very differen way of thinking about computers, and in general it is a harder way. In other words this will come down almost entirely to the tools for development. That and massive training of developers. (most of which will need a lot of adjustment to the new platform.

Re:if it sounds too good to be true..

[RicktheBrick]

IBM develops inexpensive new supercomputer than sells personal computer business to a chinese company. Why would they do this if they could use Linux and this chip and virually become a monopoly of the pesonal computer world(both hardware and software)?

Re:if it sounds too good to be true..

[zogger]

Because IBM is an R&D and service company mostly, or it looks like they are headed that direction eventually. They can make better profit margins by just designing then licensing out the tech. By concentrating on their core missions they can maximise ROI, and leave the headaches and drudge work of mass production and marketing of consumer level stuff to some other company, and still get paid well for it. Granted, you get a higher gross income with being the manufacturer, but you get a better net income by just licensing and developing.

At least it looks that way to me, and it's following their past business model of selling off consumer level production, like they did with hard drive manufacture to Hitachi. Whether that will be a very long term smooth move I have no idea, but in the short term it's actually making them money. Profit margins at low end retail are small, they want no part of that, too clunky for them. Fabbing the chips is a different story, they need to be able to have a place to build what they R & D, so in that sense its logical for them to do that,and get that aspect subsidised by licensing and direct sales (saves them research costs long run) but after that point it's just manufacturing vacuum cleaners or blenders, they don't want to, and that's all PCs are now, just another consumer appliance.

Re:if it sounds too good to be true..

[caswelmo]

"...massive training of developers."

I wonder, will this be an opportunity for US programmers to increase their worth and regain an edge? Or will this just be another area that southeast Asians can quickly adapt to and grab some more outsourcing jobs?

Re:if it sounds too good to be true..

[rjshields]

..it probably is.

Indeed. From The Article:

We have never seen a leap in performance like this before and I don't expect we'll ever see one again..

Cell is going to turn the industry upside down .. nobody has ever produced such a leap in performance in one go and certainly not at a low price.

With this encouragement, I'll bet the fanboys are foaming at the mouth already, just like when Apple announced the fastest PC in the world.

Re:if it sounds too good to be true..

[RichardX]

haha.. yeah, I remember that. Weren't there going to be PS2 guided nukes or something as well?

For an interesting look back, google around and find some pre-release articles and "screenshots" from current or previous consoles. Especially interesting are the PS2 and the Nintendo N64 - both were hyped to insane levels. Some of the "This is what the graphics will be like!" shots from the N64 were CG renders that still aren't matched in the most powerful PC or console games today.

Re:if it sounds too good to be true..

[Anonymous Coward]

It sounds like it could also pave the way for lazy developers. Like having wristwatches in the future that have cell processors but say "REQUIRES PLAYSTATION 3 IN VICINITY."

Re:if it sounds too good to be true..

[yasth]

Don't count on it. Most of the research is in the big supercomputer using countries (i.e. US + Japan, and to some extent W. Europe) that much is true. But, textbooks will be made in English, and the trick will be to get the developers when they are learning the stuff, not as a reallignment of skills. Since the SE Asia is doing more degree churning, I don't think it will work in the US's favor. I mean Textbooks in English are already used to teach the CS curiculem world wide.

Oh well look up Linda Tuples if you are bored that is what an APULET seems to be.

Re:if it sounds too good to be true..

[Anonymous Coward]

Developers who made Transformers for PS2 said they had scenes of over 35 million polygons running at 60fps with a huge draw distance. PS2 is very powerful.

Re:if it sounds too good to be true..

[G-funk]

No kidding... from TFA:

"If over clocked sufficiently (over 3.0GHz) and using some very optimised code (SSE assembly), 5 dual core Opterons directly connected via HyperTransport should be able to achieve a similar level of performance in stream processing - as a single Cell.

The PlayStation 3 is expected to have have 4 Cells."

My bullshit detector (the shenanotron) asplode. A $500 system available inside 18 months is going to be as powerful as 40 current 3 GHz opterons? I, erm, think not.

Re:if it sounds too good to be true..

[be-fan]

It doesn't sound too good to be true, if you compare Cell to other stream processors, instead of general purpose CPUs. A GeForce 6, for example, can sustain 40 gigaflops, compared to 3-4 for a G5 or P4. Given the constraints graphics chip designers work with in regards to process technology (specifically, GPU circuits cannot be hand-optimized, because of the 6-month turnaround), achieving 10x that doesn't seem that unrealistic.

Re:if it sounds too good to be true..

[be-fan]

There were stories about the US trying to stop G5s too because they were "super-computers". It just had to do with an outdated guage of what constituted a supercomputer for export purposes.

Re:if it sounds too good to be true..

[gl4ss]

ah but the author claims that it would translate directly into all kinds of general usage, that there wouldn't be any bottlenecks of that sort.

there's some fast dsp's out there too.. but that doesn't mean you could just slap one on a workstation and it would then 'smoke' everything..

Re:if it sounds too good to be true..

[snorklewacker]

For example, the system could run interactive movies that blend Hollywood's expensive production systems with intricate story lines and characters who interact realistically with the user.

Yunno, this is doable now. All it takes is GOOD WRITING AND ACTING, and a storyline that isn't on rails. Jesus, FFX was like watching an episode of yu-gi-oh, except I think yu-gi has better voice acting. Emphasis on the watching

Re:if it sounds too good to be true..

[be-fan]

Of course, the author is editorializing, like most journalists do. However, the editorializing on the part of people not affiliated with Sony or IBM shouldn't reflect negatively on Cell.

Re:if it sounds too good to be true..

[DeadScreenSky]

The "interact realistically" part is the problem. That kind of AI would be hard even if you had hardware 10 times the power of the PS2. They really were pushing the PS2 as being capable of film-quality visuals, too. None of this is really possible yet, especially with a nonlinear storyline.

(I don't disagree with you about what videogame storylines need more of - it just isn't easy, and that still wouldn't put us where Sony promised.)

Re:if it sounds too good to be true..

[DeadScreenSky]

Ratchet & Clank 1's levels had about 500k polygons total in each (reference found here, free login required). Assuming you see about a third of those polygons at all times (which is ridiculously over-the-top, especially since the game used an advanced Level of Detail system), and that maybe enemies double that polygon number (again, no way is that a realistic number), you still only get around 15 million polygons a second. Impressive, but a far cry from what Sony promised! (And more realistically, it is probably closer to half of that.)

I don't think things were particularly bumped up for the sequels, either. If anything, the LoD system was probably optimized further than anything else, decreasing polycounts. There is only so much optimization you can do in one year of development!

I think the Ratchet & Clank games do look great, don't misunderstand me here... But Sony lied about the PS2 hardware's capabilities, big-time.

Re:if it sounds too good to be true..

[binary42]

Actually you are right. The quote seems acurate. But... The game actually has many dynamic objects added to the level during game play (charater, enemies, sfx, gui). This brings the total up by a bit. Just remember the diff. between 75M simple polygons and 15M textured and shaded polygons with other efects shuch as alpha blending. The LoD is quite agressive on most games. What I was getting at was that the game looks great! The PS2 is "enough" to pull of current generation games. While it doesn't beat XBox (which, btw, also has highly inflated specs) it get the job done.

The high point that cannot be reached. You can get about 35 million vetrices per second (~40 is probably max if you use a few more tricks). I know this is a fact as I have done it myself. This is very low. But using triangle strips and fans you can get close to the above number. 15M is a bit low for an estimate of raw poly power as the 75M was supposed to represent.

All in all, I think we have rightly been mad but that does not stop the PS2 from being capable.

Re:if it sounds too good to be true..

[binary42]

Let me clarafy. I used preview but... I had an after thought. Basically the PS2 can do 35M flat or simply textured polys per second. The system is tied up doing this so if you add enough other things then you slow down a bit... plus you need steady frame rates. The IOP takes most care of most things regarding controll and other IO (sound etc.. thus the name ;) ) but the EE will have to take the brunt of things like animation and physics (its quite a bit better than the IOP for this stuff). So... a really well written engine *could* get close to 25M polygons per sec.

Re:if it sounds too good to be true..

[Anonymous Coward]

While I agree, he does imply that Cells would be beneficial for all apps, the article is mainly concerned with graphics, video, and scientific applications-all of which are ideal for this sort of processing.

Re:first post!

[SlimFastForYou]

Judging by your user name, it seems you are using outdated technology. x86 can no longer get the first post!

What always confused me

[hyu]

Something that has always confused me in gaming consoles is that, despite incredibly powerful hardware (processors, graphical chips, etc.), the system developers seemingly always neglect to put in enough RAM for most games to perform to their potential. Many PC ports often have portions compromised due to the lack of RAM, and system speeds also suffer because of this.

Seeing how RAM is increasingly becoming cheaper, is it possible that new systems like the PlayStation3 might be able to provide RAM that actually allows games to reach their potential along with this new cell hardware?

Re:What always confused me

[FRAGaLOT]

Actually I find the opposite to be true. Take for example an Xbox, which is basically a PC from about seven years ago. (Sub gigahertz P3, 64megs RAM, GeForce3 video)

But it plays all the popular games of today's PC with little to no lag. Where as you need a very high end PC to play the same game!

This is mostly due to the fact that the architecture with the video is more direct, than it is on a PC. There's no AGP bus, or any bottle neck to access video ram. It's more direct which is probably why an Xbox can perform as well as a current PC rig.

But then an Xbox is only running at 800x600. LOL

Re:What always confused me

[Anonymous Coward]

But it plays all the popular games of today's PC with little to no lag.

Holy fucking shit! That's the god-damn dumbest thing I've ever heard. Of course it plays the latest games because the latest games ARE BUILT TO BE PLAYED ON IT!!!! You expect XBox developers to write games that won't perform on the system?!?

The cross platform games usually have the graphics cranked up on the pc side to make use of the better performance. Just because a game runs on both doesn't mean that the XBox can perform as well as a PC. The games are either dumbed down for both, or specialized for each platform: some graphic features for the XBox vs memory + hd space + resolution on the pc for example.

Re:What always confused me

[Anonymous Coward]

Right. and they dont program pop ups into xbox games eh?

Re:What always confused me

[Anonymous Coward]

Less than 800x600, actually.

The XBox cannot render to a true VGA signal.

It also runs a celeron processor though it is a pentium 3-generation celeron.

It's pretty much a halfway between a geforce3 and a geforce4, as well. There are some parts of the GPU that are unique to the xbox.

And really, think about this:

All the developers that made their PC / Xbox games, had to make a special version for xbox, generally. This version included less of the detail that the PC version did.

See: Doom3.

Re:What always confused me

[metricmusic]

or they dumbed down the pc version. See: Thief 3 and Deus Ex 2. oooohhhhh hthat makes my blood boil. Deus Ex, Theif 1/2 were some of my most loved pc games. I was so eager to play their next installments. I was so disappointed by what came out.

Re:What always confused me

[John Betonschaar]

But then an Xbox is only running at 800x600. LOL

Actually, it is running at 720x576 (a PAL XBOX that is) but I don't see why this is so funny, because that's just the resolution of a PAL TV. Having a higher framebuffer resolution would probably only decrease the output quality when displayed on a normal television.

That said, if you have an HDTV, the XBOX can output at 1920x1080i...

Your sig is mine

Re:What always confused me

[Anonymous Coward]

You mention the pal xbox, then you mention how great the xbox outputs to a HDTV.

Well I happen to own a HDTV and I own a pal xbox.

Too bad the pal xbox is incapable of HDTV mode.

LOL again.

Re:What always confused me

[Troed]

Either chip your box or use a software exploit, switch it into NTSC video mode (you can do that without having to change the game region mode - your PAL originals will still work), connect a component videocable, enter the Microsoft Dashboard, select the HDTV resolutions you want from the list under Video - Settings. Play in HDTV.

... yes, all this - and you can still play PAL originals in higher resolution on XboxLive - as long as the chip is off (or the software exploit not used) afterwards.

Re:What always confused me

[Gordonjcp]

Take for example an Xbox, which is basically a PC from about seven years ago. (Sub gigahertz P3, 64megs RAM, GeForce3 video)

I would *love* to know how you had a P3 and a Geforce 3 in 1998. I had to make do with a brand-new top-of-the-range PII-350 with a Matrox Mystique and SLI-ed Voodoo 2s. Did you pull the Geforce through one of these parallel universe wormholes?

Re:What always confused me

[LarsWestergren]

Take for example an Xbox, which is basically a PC from about seven years ago. (Sub gigahertz P3, 64megs RAM, GeForce3 video) But it plays all the popular games of today's PC with little to no lag. Where as you need a very high end PC to play the same game! This is mostly due to the fact that the architecture with the video is more direct, than it is on a PC. There's no AGP bus, or any bottle neck to access video ram. It's more direct which is probably why an Xbox can perform as well as a current PC rig.


Well no, not exactly. The reason console games don't suffer from lag is that unlike a PC, the hardware specs are not a moving target during development. Developers can optimize textures, audio, algorithms etc with a specific platform in mind. This makes it much easier to create content that you know won't overwhelm the machine.

Compare this with a PC developer. They have to estimate the time it takes to develop the game. Then they have to estimate the average gamer hardware and the cutting edge gaming hardware at the time the game is released. They have to take into consideration at the very least processor speeds, main memory size and speed, graphic card speed and memory size.

If the developers overestimate, the game will be unplayable (when the first System Shock came for instance, I remember reviewers writing "You actually need a PENTIUM to play this game, it's insane!"). If they underestimate hardware or take too long, they will be killed by reviews complaining about "outdated graphics". Oh, and preferably there shouldn't be any problems with any special configuration.

This is extremely difficult to achieve. Half-Life 2 for instance was praised for the fact that it managed to scale its graphics so it was playable on low end yet good looking on high end machines. However, some people experienced stuttering of audio as levels started, this was even more noticable in Vampire: Bloodlines, a great game that uses the HL2 engine. I think this had something to do with the hard drive loading textures or level geometry (I noticed it especially when loading the huge LA Downtown level in Vampire, sound was stuttering for 10 seconds after level loaded). People with fast hard drives, especially those that chose or had to chose low resolution textures didn't suffer from it as much as those with graphics cards with lot of memory for high resolution textures and comparatively slow hard drives.

So, the interaction between the many different hardware configurations on PCs makes it difficult to optimize and that is what causes the lag, not the lack of any AGP bus or anything. A console developer can test on just on one console and be fairly certain it will run the same on all target machines.

Re:What always confused me

[John Betonschaar]

Too bad the pal xbox is incapable of HDTV mode.

Every XBOX is capable of HDTV mode...

LOL again.

I guess so...

Your sig is mine

Re:What always confused me

[LarsWestergren]

One more thing I realized I should have put in the above post, the XBox does NOT have the same performance as a high-end or even mid end PC of today. Textures have to be of lower resolution, this is easily seen for instance on Knights of the Old Republic.

Less main memory on the XBox also mean smaller levels or less content on each level. Unfortunately the cost of creating different size or content for two platforms is so expensive time-wise most developers just choose to design for the XBox and let PC users deal with lots of loading screens. This can cause negative reactions for PC owners, see for instance Deus Ex: Invisible War and other games that are percieved to have been "dumbed down" for console hardware (or maybe console owners...).

Re:What always confused me

[Anonymous Coward]

That's because RAM is way more expensive than floating point multipliers. Consoles will never have "enough" RAM because you can never have "enough" RAM.

The problem is PC designers who are incapable of limiting themselves to a design that will fit on the console that they make ALL THEIR MONEY ON. It's simply bad design.

Re:What always confused me

[Oddly_Drac]

"the system developers seemingly always neglect to put in enough RAM for most games to perform to their potential."

Rambus was expensive at the time.

"PlayStation3 might be able to provide RAM that actually allows games to reach their potential along with this new cell hardware?"

Rambus is much cheaper now.

Re:What always confused me

[DeadScreenSky]

That's kind of funny. From my perspective, games like KOTOR or Deus Ex: Invisible War on the Xbox are too limited because of their insistence on programming it as if it is a PC game. :D Lots of load times all over the place instead of a streaming universe (because you can always buy more RAM or a faster hard drive), no real attention to scene polycounts (because you can always get a better videocard), etc. (It isn't a surprise that both of these games are built on older game engines that were created for the PC.)

The Xbox and PC are certainly not equivalent at running the same games. It isn't a big surprise that a game designed for the PC will look better than a console port. But when you program specifically for the Xbox hardware's strengths, some of the games easily meet or even exceed current PC visual quality. Look at Ninja Gaiden, Amped 2 (which can run at 720p), DOA2U, Rallisport Challenge 2, or Conker: Live & Reloaded. All of these easily stand up to the best visuals you can find in current PC games. (Hell, look at GT4 on the lowly PS2 - no PC racing game comes close in graphical quality.) Being able to tailor your art assets to a powerful standardized platform makes up for a lot of excess videocard or CPU hardware.

(Obviously all that ends up really mattering is the games, of course.)

Re:What always confused me

[Deliveranc3]

Xbox provided hard drive which solved the problem... Anything else is absurd.

Re:What always confused me

[chaoaretasty]

Actually the reason that Deus Ex Invisible War was so limited is because they wanted it to do well on Xbox, the game is a shadow of its prequal, taking out all the things that made it special as opposed to "oh, another FPS".

Re:What always confused me

[Anonymous Coward]

umm, wow, what can I say? You are obviously the master of all computing problems. . .

Now for something completely different:

Programs (games) don't actually execute from the hard drive. The hard drive is *too slow* which is why PC's, and consoles, and every computing platform known to man, has system RAM. You copy stuff from the hard drive into ram when you load the program/game, and actually use it from the RAM.

The hard drive in the xbox gives developers lots of room to store save game data, and maybe to build temporary files when needed. But, when you actually go to run the game, you need RAM. The more RAM you have, the more stuff (3D models, textures, animations, game logic code, game state data, etc) you can have fast access to.

Most of the consoles have very limited RAM, which means in most games, you either have small 'levels' (maps, whatever - the game is broken up into chunks anyhow) that get loaded and unloaded when you go from one level to the next, or one area of the game to the next. More RAM would mean game developers could do two things: either have larger game areas loaded, so you would have fewer loading screens, or have more detailed areas/effects/game data. Or both.

Having a hard drive does not solve that problem.

Re:What always confused me

[DeadScreenSky]

I think that a much bigger factor is that Warren Spector wasn't the lead designer anymore. A game like DE simply isn't something that can reasonably be accomplished by someone new to that position. The game has all the hallmarks of an inexperienced dev team (in particular, note the tonal and pacing issues).

Anyway, hardware really didn't affect the game like some people pretend it did. The streamlined gameplay was because Harvey Smith and his team wanted it that way (the Xbox has certainly seen plenty of more complicated gameplay systems than Deus Ex 1!), and the vast majority of them would have occurred even if the game was PC only. I disagree with the decision as well, but the devs wanted the gameplay to be simpler and more focused.

Most of the engine limitations were simply because they chose poor technology - the hacked-up UT2k3 engine didn't scale or perform well on the PC, either. Lots of Xbox games feature huge levels with minimal or even non-existent load times (see Riddick, Halo series, Ninja Gaiden...).

(And I would point out that the original DE was pretty spotty when it came to tech as well. Very slow on most systems when it was released, without particularly nice graphics to compensate. I suspect they just donn't have the calibur of 3D programmer required...)

And as much as I love DE1, it really didn't leave room for a sequel. Near-future stuff works fine, but once you get close to or even past the Singularity, it is almost impossible to create a realistic or interesting setting (as any human just wouldn't be close to where the real action is). Most of the truly interesting conspiracy theories were already dealt with (seriously, you already used the Templars, the Illuminati, and Majestic 12). Most of the interesting future tech was used (nanotechnology especially, though I do think Invisible War expanded on it in interesting areas). And you couldn't reasonably expand on the cyberpunk theme too much, because the world had already been pulled back from the brink in the original. (The globalization issues that Invisible War brought up was a good attempt, but that is really hard to address in a FPS - there are no real masses, you understand?)

Re:first post!

[Anonymous Coward]

Haha you got robbed at 3am in the morning...!

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

WTF are you talking about?

Logical Progression?

[grimsweep]

Is the integration of multiple processors into a single 'unit' the answer to the alleged woes of pushing a single processor as fast as it can go? I seem to be seeing more and more consumer-oriented solutions that involve multiple processors. Is this simply due to reduction in costs, or is it destined to become the norm? Some technologies (i.e. Intel's hyper-threading) appear to be a prelude to the latter.

remember ps2 technology

[Anonymous Coward]

remember before it was released, how powerful they said it was, so powerful and revolutionary it may not be exportedto some country, as it could be used to guide missile and stuff...etc...

whell, usual PR stuff, sony is very good at it, what amaze me is that it works again and again....people never learn they say...

Re:remember ps2 technology

[Anonymous Coward]

"could be used to guide missile and stuff...etc..."

A whole lot of people are laughing their asses off at you moron.

Just a stupid fanboy who shot himself in the foot. Congrats!

Re:remember ps2 technology

[Anonymous Coward]

i am not sure i fully understand you're answer...
but just in case, i never owned a ps2 console (no ps1 too). my last and only console to date was a dreamcast (long ago...)

so i don't really think i am a sony fanboy...i am more on computer stuff...

but maybe i missed something?

Re:remember ps2 technology

[Ash-Fox]

My last console was a NES!

Re:remember ps2 technology

[Agret]

I'm not sure if you realise a 386 can guide a missile.

Re:remember ps2 technology

[Anonymous Coward]

I'm pretty sure a 6502 can guide a missile.

Re:remember ps2 technology

[hunterx11]

Oh my God! We had better make sure the terrorists don't get their hands on an NES.

I also hear that they'll be able to render Tron-style graphics in real time!

Re:remember ps2 technology

[Anonymous Coward]

You don't even need a digital computer to guide a missile.

I'll believe it when I see it

[Anonymous Coward]

I'll believe it when I see it. Sony made outrageous claims with the PS2 in the year or so before launch, I see no reason to believe this will be any different.

On paper an Emotion Engine was supposed to destroy everything, but achieving maximum throughput was difficult and other contraints such as I/O and memory hampered performance. Programmers had to learn a very different way of programming to make full use of the processor and it's two vector units.

A Cell might be a killer chip on paper, but real-world hardware with I/O latency and memory contraints will bring things down to a more reasonable level. Don't forget that multiprocessor programming is *hard*.

Hopefully, developing software for Cell chips will be easier then the early days of the PS2, Sony has already said as much a few months ago.

Re:I'll believe it when I see it

[Anonymous Coward]

The PS2 is almost 5 years old (March) and it's still able to compete with the XBox which is what... 2 years old? Grab a clue foo, it blew everything away in 2000 and it's still pretty damn good today.

Re:I'll believe it when I see it

[Anonymous Coward]

Don't forget that multiprocessor programming is *hard*."

Let me guess, you heard this from that 'console' programming 'god' J Carmack?

But multiprocessor programming is difficult, as anyone who has done it will tell you. Some things are easy to parallelise (ray tracing springs to mind), but most things are not.

True, you can usually get a speed up on 2 CPU's, but unless you are careful then synchronisation & other communication overheads will kill your performance. Adding processors can actually slow you down.

I am not a game programmer, but I do write 3D renderers for a living.

Re:I'll believe it when I see it

[Anonymous Coward]

Us console programmers have been writing code for concurrent multi-processor systems for a long time. It's not even worth discussing.

However, there are a lot of dimwitted x86/directx coders shitting their pants now that clock speeds are reaching a wall...

Re:I'll believe it when I see it

[Ash-Fox]

So does my amiga built in the early 90s, your point being?

Re:I'll believe it when I see it

[Anonymous Coward]

Funny how this ultra powerful core, much faster than four dual core Opterons or any Nvidia graphics card, is still going to need an Nvidia (or did they go with ATI?) graphics card in the PS3 to do the graphics. This article is just hype. Vector stuff is good, has potential, but this isnt going to be that good.

Re:I'll believe it when I see it

[Anonymous Coward]

Us console programmers...However, there are a lot of dimwitted x86/directx coders shitting their pants

Nice to see console programmers have the same maturity as console users...

Re:I'll believe it when I see it

[Anonymous Coward]

When your market is ten times the size, you can talk all the shit you want.

Re:I'll believe it when I see it

[pommiekiwifruit]

Sony claimed that the PC world would not catch up with the PS2 for many years (was it 5 or 10 I forget). The Xbox (and Gamecube) showed that claim up with real-world apps.

Re:I'll believe it when I see it

[Anonymous Coward]

"Hopefully, developing software for Cell chips will be easier then the early days of the PS2, Sony has already said as much a few months ago."

I wouldn't bet on it...

From TFA
"Cell uses the same approach, these is no cache in the APUs, only a small but very fast local memory is present."

So there is a small amount of very fast ram. Either this interacts with a larger amount of slower ram - in which case it's a cache - or it doesn't, in which case there isn't very much ram. I understand that dealing with swapping in the ram on the PS/2 was a source of much difficulty.

"There is another reason I don't think Nvidia or ATI will be able to match the Cell's performance anytime soon. Last time around the PC rapidly caught up with and surpassed the PS2, I think it is one of Sony's aims this time to make that very difficult so, as such Cell has been designed in a highly aggressive manner."

As opposed to when Sony were making the PS/2.
Sony Exec: "Hey, guys, should we make this chip in a very aggressive manner, so that the PC can't catch up?"
Sony Boardroom (together): "Naaa..."

Re:I'll believe it when I see it

[MORB]

If the repartition and interconnection details of the cells is indeed done automatically, it should make things much easier.

However, you'd still need to program in a different way than today, separating and pipelining things into separate software units that could run on different cells.

I guess having a good software component model that interface nicely with current programming languages will be necessary to relieve programmers from the details of the communication, exchange data formats and such, between the software cells.
Hopefully, they are working on these things aswell as on the hardware itself.

Re:I'll believe it when I see it

[hazee]

You're not wrong to say that developing parallel algorithms in general is hard. It wouldn't be much fun to try and code something like Word to make full use of x processors, say.

However, these days, most games are all about graphics and, by a happy co-incidence, that is one area that *does* work really, really well with parallel processing.

Just look at all the ray tracing systems that will happily soak up as many processors as you can throw at them - whether that be a single box with multiple processors, or a Beowulf cluster or the like. At the simplest level, you can just divide up the image into sections, and have each processor work on its own section.

So while parallel programming is hard, a games console might just be the ideal place for it.

Will they be able to match the total theoretical system performance? Probably not. Will they be able to make good use of all those parallel processors? Hell yes.

Re:I'll believe it when I see it

[Radius9]

I'm a professional console programmer, and I call bullshit on your statement. Programming multi-processor systems is not only hard, but doesn't always work particularly for games. Most of my rendering is dependent on, for example, the physics system. The physics system is dependent on my AI system. In my AI system, although it seems like I might be able to parallelize it easily because you have a bunch of different characters, its not, because the characters all interact with each other. I can't run the physics and collisions and until I've run the AI, and all the physics objects also interact with each other, once again breaking parallelization. The rendering can be started once I'm done with AI and physics while I start on the next frame, but that is assuming that I have enough RAM to run both at once, something that's unlikely given the cost of RAM. So being able to run things in parallel isn't really that useful most of the time, as anybody who has worked on the PS2 can tell you.

Re:I'll believe it when I see it

[ciroknight]

one *tiny* question: why the hell does your physic's unit require any relation at all to the AI system? Just because you'd have to wait on the AI system doesn't mean that the physics system can be running environmental type data until it's ready to process character and object interaction. I think it wouldn't be too bad if you could schedule everything properly, which would be just as much of a task. You're right that not all aspects of the game need to be paralleled, but those that can be, should. Even if that means having to go back to platform style coding where all the graphics are coded and rendered in layers, and then overlayed on the final output. (like, rendering characters out of environment, etc. may not be the best of examples, but use your imagination.)

Re:I'll believe it when I see it

[hitmark]

allso, a lot of computers multitask, but every task is a single string. now toss every active task to each own apu (or whatever that article called em) and you have a interesting setup ;)

Re:I'll believe it when I see it

[x40sw0n]

of course they made outrageous claims, that's what maketing people are for. Did they make a 'killer app'? Yes. there is no doubt that the ps2 is THE dominant console, even if it is an aging unit. here in the US you can make the claim that the xbox is better but the numbers don't lie: units are higher on the ps2 even here where the xbox is strongest. in Japan the xbox has never even hit its target numbers, in fact they barely made 10% of their projected numbers in Japan. And the thing is, if it doesn't fly in Japan it won't fly in the console market. Japan is what killed the Dreamcast after all. is the xbox more powerful? maybe, but not by much. can a pc do what a ps2 does? not yet, and here we are 6 years later... sounds pretty close to on the money to me. when a pc can succesfully emulate a console, then it has surpassed it. Dreamcast is doable, so is the Playstation; they are close on the ps2 and just as close on xbox. The other thing is this; developers for a console need one really important thing: support. Sony has bald faced stated that this console will be much easier to develop content for. This statement says two things in my mind: 1. They recognize that they unintentionally made development difficult on the PS2; 2.They have a plan in mind to avoid it (I have had rumors of special developer kits already shipping, though this is all software based, running on pc's, which means that a relatively standard architecture is being used on some level. possibly it is a self-coding development tool that ports code to the cell processes, or they are modelling their cpu-software interface on clustered computing projects. Maybe there is just a simple interface layer that is involved (hardware) that makes the multiple cpu's appear as one RISC chip to the software... after all in the early development materials they did say that it could 'share' workload with other Cell devices on a network... the materials made it sound as if all the process workload was divided at hardware level not software. maybe coding for paralell processes won't even be necessary. could be we are taking SMP assumptions into a MMP world?

-x40s0wn

Re:I'll believe it when I see it

[zardo]

Yes, wasn't this the fate of the sega saturn? Or was it the atari jaguar? It used multiple processors which made it a fast and cheap system, but developers steered toward the ease of programming on the N64

Re:I'll believe it when I see it

[forkazoo]

Both had odd architectures. The Genesis also had two CPU's. But, what really killed them is that they weren't two of the same! They had different chips that did different things. Imagine having a Power PC and an Athlon is one box, and trying to write an efficient program that would load balance between the two! not fun!

Re:I'll believe it when I see it

[cbr2702]

So there is a small amount of very fast ram. Either this interacts with a larger amount of slower ram - in which case it's a cache - or it doesn't, in which case there isn't very much ram.

Not really. A cache sits between the processor and the memory and completely transparently to the program stores often-acessed memory to improve overall response. The local memory is different because it is completely distinct from main memory. If the programmer knows that a certain value will be used a lot (but not enough to merit a register) it should go in the local memory. This shifts some work to the programmer or compiler but should result in a faster program overall.

Re:I'll believe it when I see it

[evilviper]

Yes, wasn't this the fate of the sega saturn? Or was it the atari jaguar? It used multiple processors which made it a fast and cheap system, but developers steered toward the ease of programming on the N64

The Sega Saturn used dual processors, and was nearly a clone of top-end Sega arcade systems. Unfortunately, it was terribly hard to program, so only in-house Sega titles were developed to utilize the full potential of the device, such as Virtua Fighter, while other titles were only using half the performance of the system.

It was not, however, cheap.

The Jaguar was cheap and was hard to program for (primarily because of bugs in the cheap hardware), but it could hardly be considered a powerful system. It was well ahead of NES/Genesis, but it came in late in the game, when they were both well established, and the Playstation/Sega Saturn were just around the corner.

Re:I'll believe it when I see it

[Surye]

It's able to compete based on allegance and ... game library allegance. It doesn't stack up otherwise. PS2 games are fucking ugly compared to any XBox game, even as a port of the PS2 game.

Re:What's that? Microsoft isn't supporting it?

[Peter Cooper]

Well if even a Playstation 3 is going to be the equivalent of 20 Opterons, I imagine you could emulate an x86 in software (a la Virtual PC) and it'd still be as fast as, oh, a few Opterons? :)

I don't know if I can take this article seriously. A games console due out within the next year or two is going to be as powerful as 20 of our current top of the range chips? I'm not buying it.

From TFA...

[Trimbo2]

"Some will no doubt be turned off by the fact that DRM is built into the Cell hardware. Sony is a media company and like the rest of the industry that arm of the company are no doubt pushing for DRM type solutions. It must also be noted that the Cell is destined for HDTV and BluRay / HD-DVD systems, any high definition recorded content is going to be very strictly controlled by DRM so Sony have to add this capability otherwise they would be effectively locking themselves out of a large chunk of their target market. Hardware DRM is no magic bullet however, hardware systems have been broken before - including Set Top Boxes and even IBM's crypto hardware for their mainframes."

Re:From TFA...

[Anonymous Coward]

I think that I stick to PC+Bluyray burner when recording HDTV content...

Re:From TFA...

[DigiShaman]

In other news, over 90% of the CPU cycles will be used to encrypt and decrypt the DRM content to prevent hacking.

It's Sony folks, what did else did you expect?

Re:What's that? Microsoft isn't supporting it?

[xylix]

Unless the thing has an x86 emulation layer, it's dead in the water in regards to the PC market. Even Apple, with their much vaunted G-chips have to emulate the x86 hardware so that users can run their Windows programs.

Huh?? what are you smoking? Since when does Apple emulate x86 harware? Perhaps you are confused by the fact that you can buy Microsoft Office for the macintosh, or run Internet Explorer. Heres a news flash - they aren't emulating the x86, they are native mac code.

Most likely you are confusing one of two different things. Back in the OS X programs can be either Cocoa (the new way) or Carbon (the old way) apps. But that isn't really emulation. What is emulation is that Mac OS X can run OS 9 by emulation - nothing to do with x86 here.

Or you can buy Virtual PC, which is now owned by Microsoft. It will allow you to emulate x86 to run Windows (or Linux) and associated. But note - this isn't Apple, and it isn't something that Mac users "have to do".

Lastly, why on earth would you HAVE to be able to run windows programs in order for a Playstation processor to be successful? Last time I checked PlayStation was still wildly successful? More so than MS Xbox I think.

Can this be taken seriously?

[Anonymous Coward]

Quotes from article:

"GPUs will provide the only viable competition to the Cell but even then for a number of reasons I don't think they will be able to catch the Cell."

Did this guy forget that NVidia is designing the GPU for PS3? If Cell is so almighty, why does Sony uses NVidia GPU instead of using more Cells for graphic prosessing?

"There is another reason I don't think Nvidia or ATI will be able to match the Cell's performance anytime soon."

Of course, Cell based products won't be available anytime soon either. According to the current rumors, PS3 will be available in Japan in Spring 2006 and elsewhere in Autumn 2006. One and half years equals a generation in the GPU world...

I love this kind of articles where some future products are compared against current ones and declared as a clear winners...

Re:Can this be taken seriously?

[Anonymous Coward]

You asked why then is Sony using Nvidia's GPU?

Well, the answer is 2parts:

1. is because for certain operations the GPU is going to be faster, and

2. (and the most important) The Cell CPU doesn't include a built-in video controller... duh.

Re:Can this be taken seriously?

[binary42]

Quite simple. The PS2, for example, has a GPU (called the GS -- graphics synth.). The pipe that usually feeds the GS in the PS2 is actually one of the vector units that is part of the EE (VU1 I think). This vector unit provides an intermediate processor for primitives sent to be rastered by the GS. Most of the time this kind of stuff is done by the GPU on a traditional architecture (i.e. XBox) but the PS2 extracts it. This makes PS2 very flexible but harder to program (their tools have become better over time -- like their Vector Assembly tools important to higher quility games we've seen lately). The Cell on the other hand will have more vector power and MUCH more bandwidth. If all your processing is done and you just need the darn thing to be drawn then you need a good rasterizer and that is just what nVidia will probably give them.

Re:Can this be taken seriously?

[jovetoo]

First of all, as another post says, the GPUs contain a video controller, DAC and so on. Second, the Cell will still be able to accelerate graphics performance by doing all kinds of vector pre-processing. Last, it will be a lot more easy for software companies to build PS3 games fast if they have somewhat the same computing/graphics environment as on a x86. Reasons enough, I think.

But what struck me most is that you seemed to have missed the whole point the authors seeks to make. Yes, Moore's law will double the performance of the GPU within 18 months. So? It still does not give them the raw processing power of those Cells. Nor the scalability. (Damned! These things will be in you TV, your DVD, your stereo and they all cluster...) If these Cells really become low cost chips, I seriously doubt x86 will survive.

Re:Can this be taken seriously?

[mormota]

I guess what the author means, is that GPUs can be a viable competiton to Cell on the field of general purpose vector processing, and not the other way arond. No?

Re:Can this be taken seriously?

[EhWhat]

> Did this guy forget that NVidia is designing the GPU for PS3? If Cell is so almighty, why does Sony uses NVidia GPU instead of using more Cells for graphic prosessing?

Why are you so sure that Nvidia are using their own hardware tech? It makes almost as much sense to get NVidia's graphics optimisation and other intellectual capital creating streaming apulets on Cell

Re:Can this be taken seriously?

[Anonymous Coward]

If Cell is so almighty, why does Sony uses NVidia GPU instead of using more Cells for graphic prosessing?

That's not what he's saying, if you take the quote in context. He's saying that a PC using the GPU for general-purpose computation won't beat the Cell at the same computations.

What he's specifically not saying is that the Cell is better than a GPU for doing graphics. The GPU is a special-purpose chip for graphics, so of course it's going to outperform something more generic, on cost if on nothing else.

Re:Can this be taken seriously?

[Anonymous Coward]

There is no such thing as "graphics optimisation" capital. nVidia do chip design, with software libraries (which they are BAD at developing) as a secondary concern. The optimisation they are good at is chip optimisation. The software optimisation that would be required for doing fast software rendering does not exist at nVidia - just look at their drivers.

Re:Can this be taken seriously?

[Anonymous Coward]

>Why are you so sure that Nvidia are using their own hardware tech?

Because they say so, where have been for last few months? :). For example here is an interview of David Roman from NVidia:

"Anna (X-bit labs): From the technology stand point what will the new GPU developed for Playstation 3 resemble if we compare it to the desktop chips by NVIDIA?

David Roman: This is the next generation GPU, so after the GeForce 6 series this is going to be the next generation. So, this will be everything we have in GeForce 6 + whatever else we bring out. Obviously it will support DirectX 9, shader models 3, it will be the most feature-rich, the most powerful GPU that weve ever created actually, when it comes out."

Yes, NVidia will be designing the hardware for the PS3.

Re:Can this be taken seriously?

[Anonymous Coward]

"Yes, NVidia will be designing the hardware for the PS3."

???

Uh, no.

Re:Can this be taken seriously?

[Anonymous Coward]

"First of all, as another post says, the GPUs contain a video controller, DAC and so on. Second, the Cell will still be able to accelerate graphics performance by doing all kinds of vector pre-processing."

By definition, GPU is Graphic Processing Unit. Term GPU was used for the first time when Geforce was introduced with integrated T&L unit in 1999.

"Video controller" refers to the old VGA/EGA cards, it has nothing to do with modern GPU.

Currently, virtually all transistors within the GPU are making calculations required for graphics. DAC takes very small amount of transistors and in many graphic cards external DACs are used.

Since the current GPUs have very complex and programmable Vertex shaders, CPU doesn't need to do any vector pre-processing. In fact there have been talks about moving physics and AI calculations from CPU to the GPU!

"But what struck me most is that you seemed to have missed the whole point the authors seeks to make. Yes, Moore's law will double the performance of the GPU within 18 months. So? It still does not give them the raw processing power of those Cells."

He didn't have much proof for hits points. He didn't talk in detail about capabilities of the current GPUs, let alone the future one. His specs for PS3 are totally out of this world (come on, 4 CPU cores, 32 ALUs running at 4.6GHz within a tiny console?)

Re:Can this be taken seriously?

[mwvdlee]

You seem to think the Cell chips are equivalent to GPU's which, if you would have read the entire article you would have known, is not true.

Cell chips are somewhat like the shader units in a GPU in that they are dedicated vector processing units. In this capacity, a GPU's shader unit may compete with a Cell processor.

As you will probably understand now, this still leaves a large part of the GPU which is not covered by Cell chips, and although the Cells will be able to handle that math, a dedicated GPU will do much better.

Although anything regarding NVidia's PS3 GPU is speculation at the moment, I suspect they will just implement a "basic" GPU with all the shader units "replaced" by the Cell chips, that would seem to make the most sense from a technical point of view.

Re:Can this be taken seriously?

[Oddly_Drac]

AC's astroturf.

"Did this guy forget that NVidia is designing the GPU for PS3?"

"Both companies are jointly developing a custom graphics processing unit (GPU) incorporating NVIDIA's next-generation GeForce(TM) and SCEI's system solutions for next-generation computer entertainment systems featuring the Cell* processor."

Joint development means close collaboration; if Nvidia are the first to lever cell into their GPUs, they could nail ATI to the wall.

"Cell based products won't be available anytime soon either."

They'll be announced in March, but not on a 65nm die.

"I love this kind of articles where some future products are compared against current ones and declared as a clear winners..."

I love anonymous cowards that skim articles and speak from a position of ignorance. Or stupidity.

Re:Can this be taken seriously?

[crazy_pikachu]

I think that the PS3 and the PC are going to have a good race. The PS3 might have better graphics for a year or so but when the new technology comes out for the PC the PS3 will be left out in the dust. That is why now the computer has better graphics then the PS2 & X-Box. I think that PC will prevail though. Just because of the way we get to upgrade and the counsels don't.

The only reason that people say that the counsel has better graphics then the pc is because they are playing on a crap computer. But to any one that still says that the counsel has better graphics then the PC. I want them to come over to my house and play half life2, Doom3 and Roam Total War.

Re:Can this be taken seriously?

[hin72]

>> If Cell is so almighty, why does Sony uses NVidia GPU instead of using more Cells for graphic prosessing?

the "best" hardware in the world is next to useless if you don't have the software to fully exploit the silicon. nVidia will provide much more than just a bunch of engineers to (co)design the GPU. nVidia is also providing the whole development platform that will be available to the anyone developing games for the PS3.

Re:Can this be taken seriously?

[Anonymous Coward]

I was thinking the same thing. I also read an article at http://www.gamesindustry.biz/content_page.php?sect ion_name=dev&aid=5839 and it says (cursive is mine) :

The deal between NVIDIA and SCEI is described as a "broad, multi-year, royalty-bearing agreement," which will "encompass future Sony digital consumer electronics products," and also includes "a variety of graphics development tools and middleware" according to Sony boss Ken Kutaragi.

Early PS2 games sucked because nobody knew how to program it.

I am of course speculating but NVIDIA could also be developing an OpenGL library / driver for the PS3.

Re:Can this be taken seriously?

[AddressException]

Ummm, I think you mean 'console'.

Re:Can this be taken seriously?

[jovetoo]

My apologies for the flawed terminologie.

"Since the current GPUs have very complex and programmable Vertex shaders, CPU doesn't need to do any vector pre-processing."

The complexity of those GPUs is exactly what limits their scalability. You can only put so much transistors on a chip... does the fact that they have taken SLI from under the dust again not tell you anything?

"In fact there have been talks about moving physics and AI calculations from CPU to the GPU!"

And why do they talk about moving it there? Because the GPU is more powerfull than the CPU. Which, with Cells would not be true.

"He didn't have much proof for hits points. He didn't talk in detail about capabilities of the current GPUs, let alone the future one. His specs for PS3 are totally out of this world (come on, 4 CPU cores, 32 ALUs running at 4.6GHz within a tiny console?)"

It all depends on the price of a Cell. If these are indeed meant to be built into TVs and standalone DVD players, then they are going to be cheap.

What is more, it is the ideas behind these Cells that make them fast, not their clockspeed. Half the clockspeed and it is still has more punch than the current x86 CPUs.

They have solved the main problem of the current CPUs. Memory bandwidth. They paid for it in software complexity, but so what? All current OSses already contain an advanced memory manager.

Moving complexity to the software level makes a lot of sense. Lots easier to debug... and that makes the hardware all the cheaper. No complex hw memory mananger, no complex CPU cores. Just a set of (relatively) simple ALUs and a proven CPU core as controller.

Re:Can this be taken seriously?

[gstoddart]

I love this kind of articles where some future products are compared against current ones and declared as a clear winners...

What ... like Longhorn? Duke Nuke'em Forever? Next Year's car?

Since the future product doesn't exist, and is expected to be an improvement over current ones, you have to define it in terms of the current ones don't you?

"In the future this should be comparable to a another product which doesn't exist yet either but we can't tell you what that means since we'd have to compare to a current product" would be awfully confusing.

You need a basis to compare.

Re:Can this be taken seriously?

[Anonymous Coward]

Duke Nukem Forever is WAY better than Half Life 3!

Re:Can this be taken seriously?

[bollox4]

"Did this guy forget that NVidia is designing the GPU for PS3? If Cell is so almighty, why does Sony uses NVidia GPU instead of using more Cells for graphic prosessing?" Price? Why spend a fortune trying match the graphicly rich quality on the PC, when the hardware & development tools are already plentiful. Makes sense if you're trying to hit low consumer price point.

Re:Can this be taken seriously?

[danila]

Nope. In fact it is well-known that nVidia has a huge department optimising its drivers for current games.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

The article...describes why it is...the most serious threat to x86.

This is a line (excerpted) from the writeup. These two chips (x86, Cell) aren't even in the same ballpark, let alone market.

Re:What's that? Microsoft isn't supporting it?

[T'hain Esh Kelch]

Even Apple, with their much vaunted G-chips have to emulate the x86 hardware so that users can run their Windows programs.

You sure must be smoking crack or something! Apple has been using Power-PC chips the last almost 10 years! The Cell chip is supposedly based on PPC technologies, so Apple is actually one of the companies that might actually benefit greatly from this chip!

Apple making Windows programs only... PffftT!

Huh?

[Anonymous Coward]

In one paragraph (Games):

"The Cell designers have concentrated on raw computing power and not on graphics"

In the next paragraph (3D Graphics):

"Again this is a field the Cell was largely designed for so expect it to do well here"

So which is it???

Re:What's that? Microsoft isn't supporting it?

[popo]

"No Apps"? Try every single video game publisher in the world.

And besides, this isn't about "Office" style apps. Its about games, and more importantly: its about home media centers. I think the Windows MCE is going to have its rear-end handed to it by the PS3.

When you consider that a cell-based PS3 could have a computational power of *several times* a 3 GHz Pentium...

You have to ask, what's more likely: that Intel can get around IBM/Toshiba patents in time for Windows to conquer the living room with a faster box? (That's if they can even build a secure, stable OS with a decent UI). Or that Sony, now armed with the worlds fastest consumer-computing platform, an enormous user base and years of TiVO experience, will own the living room media center market.

If I had to bet on who builds a better media-center PC .. Sony or MSFT... I'd say its absolutely no contest. Sony would crush MSFT. They have better interface design, fewer conflicting platform goals, and they'll put a PS3 in your living room for a fraction of what MSFT could.

The architecture

[brunos]

I am not sure whether I get it or not, but it seems to me that what is revolutionary here is not just the speed, but the convenience of the architecture: what if you had just standard slots in your computer in which you could plug extra processors, graphics cards, memory, optical drive hard drive ... anything, using the same exact bus. That would make it so convenient to upgrade and to scale, that it would take over the market. Especially if you could plug in a i386 module too.

Re:The architecture

[Xyrus]

Sort of like the Amiga you mean?

They said the PS2 was going to be powerful and simple to program as well...

Let's just say I think this is more hype than fact.

~X~

Re:The architecture

[AddressException]

I don't remember the Amiga's processing power being able to be expanded infinitely.

Seems like overhype.

[Anonymous Coward]

Reading between the lines suggests to me that the article's author is getting overly excited, and extrapolating a bit too much. The whole thing, as described in the article, seems to be way overpowered and overcomplex for a games console; even if it's technically possible, I strongly suspect that a CPU with ancillary devices as given would be exorbitantly expensive.

I cannot see Sony losing so much money on their next gen console as to make this a possibility; they'd be needing far too many games purchases to make up the losses. My money's on the technology being a scaled down version of the setup described -- and when I say scaled down, I mean scaled way down. I'd estimate it's at least five, possibly ten, years before something like this is anywhere near economical enough for a gaming console.

Having said all that, I'd love to be proven wrong. But until the console comes out, we really have no way of being sure one way or the other, and I'd rather have relatively low expectations and be pleasantly surprised than to raise my expectations to Alpha Centauri (at ~4 light years :) and be majorly disappointed.

One way or the other, time will tell.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

Unless the thing has an x86 emulation layer, it's dead in the water in regards to the PC market.

If this thing can run as fast as claimed, you can have a x86 virtual machine written in Java and it will still outperform all other Wintel machines

Re:What's that? Microsoft isn't supporting it?

[metricmusic]

listen to the wise old man:

With great power comes loads of software

not a new architecture, and it's going to be tough

[idlake]

This sounds like a little PVM-cluster-on-a-chip. It also sounds like it's a pain to program and will, in the short term, suffer from the same problems that Intel's Itanium suffers from: it tries to push too much work on the compiler or software developer.

In the long term, it's nice that companies are exploring these kinds of architectures. It's not nice that they are trying to monopolize what are pretty straightforward architectural choices with patents. This may be a new CPU, but there is little that is new about having a bunch of fast processors interconnected via a reconfigurable network; these just happen to be on the same chip.

Cool!

[Jacco de Leeuw]

85 Celcius operation with heat sink

Well, perhaps "cool!" is not the correct response...

Re:Cool!

[Trimbo2]

+1 Funny btw :D

On a serious note, The cell misses the point, we don't NEED any more CPU power, what we need is existing levels of power but without the need for excessive cooling and the fan noise that goes with it (I can hardly type this over the noise of my P4 3.0) !.

Fast, Silent and power efficent is what's needed next.

Re:Cool!

[ponos]

85 Celcius operation with heat sink

Well, perhaps "cool!" is not the correct response...


It says with a heat sink only. Not with a fan!
The last chip that worked without a fan was the 486DX33 and
486DX40(I'm talking mainstream desktop PC hardware, not mobile solutions). You could probably stick a fan and get it down to
40 degress, while a Pentium 560 will produce liquid plasma and/or a fusion reaction if operated without a fan.

P.

Re:Cool!

[Lonewolf666]

If 85 degrees Celsius are the maximum allowed case temperature, this actually hints at lower power consumption than current PC CPUs.

If we have a given maximum junction temperature TMaxJ for the transistors, a total heat dissipation P and a given thermal resistance TRes between the processor die and the surface of the heat spreader, the maximum case temperature TMaxC is

TMaxC = TMaxJ - P * TRes.

If we assume the Cell's TMaxJ and TRes are similar to that of a PC CPU, a higher TMaxC suggests a lower P. PC CPUs have a TMaxC around 70 degrees Celsius these days, less for the high end models with greater heat dissipation.
Of course, this is highly speculative, and when actual specs are released, we will know more.

Re:Cool!

[hcdejong]

A PPC G3/300 works perfectly well without a fan.

Re:Cool!

[ultranova]

The last chip that worked without a fan was the 486DX33 and 486DX40(I'm talking mainstream desktop PC hardware, not mobile solutions).

I ran a Pentium-120 for a year without any kind of cooling - not even a heatsink. It worked perfectly.

Re:Cool!

[bjpirt]

The via eden processors run pretty well without a fan - they go up to 600MHz I think (see more info here)

Re:Cool!

[mbbac]

I don't know. My Power Mac G5 Dual 2Ghz is pretty quite most of the time.

Re:Cool!

[mbbac]

No, the Power Macintosh G4 Cube worked without a fan. It used a 500Mhz PowerPC 7400.

Re:Cool!

[mwvdlee]

It's called "Underclocking".

http://en.wikipedia.org/wiki/Underclocking

Re:Cool!

[MrHanky]

a Pentium 560 will produce liquid plasma and/or a fusion reaction if operated without a fan.

Now there's an idea: Power up Pentium 4 560 to start fusion reaction, and use fusion reactor to power Pentium 4 560. Voila! First self-powering CPU. I think Apple will switch to Intel to use these in their laptops: Good battery time (without battery! yay!), and so hot that you have to use asbestos gloves to touch the keyboard.

Re:Cool!

[narcolepticjim]

Way wrong -- my Pentium II 233 had no fan (not even me).

Re:Cool!

[ThousandStars]

The issue is not the chip but the overall design. One very recent computer that's received too much press on /. comes without a fan. If someone started offering Pentium M motherboards, I'm sure we'd see plenty of quiet, fanless x86 systems, and the PM is a very fast chip. Give one a gig or two of RAM and a fast HD and it'll give most P4s a run for their money.

Re:Cool!

[2nd Post!]

Well in 1999, the Apple iMac's 350MHz G3 was fanless.. So we're talking about a 10fold increase in MHz and probably a 3fold increase in compute efficiency for a 30x increase in performance, and all fanless.

Re:Cool!

[verus vorago]

My AMD64 runs fine without a fan - currently sitting at 39 C/102 F.

There are some exceptional heat sinks out there...

Re:Cool!

[stephentyrone]

strange, i haven't noticed the sound of my computer (mac) in months.

Re:Cool!

[Anonymous Coward]

I was thinking the exact same. The processors run reeally hot and there are supposed to be several of them. So imagine what 6 or 7 of these Cells and the cooling, which propably will be air, not water.
So the question here is: Will I be able to hear my own voice behind the humming of the fans? No, seriously...

Re:Cool!

[ponos]

My AMD64 runs fine without a fan - currently sitting at 39 C/102 F.

Mine too, but this is not the default specified solution. You could run any chip without a fan using a big enough heatsink. However, the simple point I raised is generally valid: every generation of AMD/Intel desktop processors needs more cooling than the previous and usually this implies a fan. I run a low-noise PC but I don't consider my components to be "default", while a fanless by default 4.0GHz high performance processor is something to note.

P.

P.S. I also run my Radeon 9800Pro without a fan. I'm now considering a fanless power supply so that I can run a really noiseless PC.

Re:Cool!

[Mycroft_VIII]

Actually someone HAS just started to offer pentium-m mobo's for desktop systems. IIRC it got mentioned here on /. about a month ago.
Though this is new enough I wouldn't expect to find them on the shelves or in shipping systems yet.

Mycroft

Re:Cool!

[Mycroft_VIII]

A good case can go a long way there. I went from a generec beige monster (FULL tower, 6 5.25 bays and 3 3.5 external) to a cooler master. While the OCD types who really work at it can do much better, the difference I notice was rather significant for me. I'm down to a quite hum I can't hear over the room fan I use to circulate air and drown oustide noises out with.

Mycroft

Re:Cool!

[Nermal6693]

The Mac Mini is not fanless.

Re:Cool!

[evilviper]

The last chip that worked without a fan was the 486DX33 and 486DX40(I'm talking mainstream desktop PC hardware, not mobile solutions).

How about a PIII 933MHz?

PIII-933 512KB FCBGA (6-B-x) 11.61W MAX

Desktop processor, and lower-power than most VIA processors which are commonly being run fanless (and will absolutely blow-away the fastest VIA processors performance-wise).

Re:Cool!

[Johnno74]

The last chip that worked without a fan was the 486DX33 and 486DX40

Not true... I just picked up an old p3-450 compaq deskpro that work was gonna chuck (for a home webserver) and its got a friggin HUGE aluminium heatsink, and no fan - just a medium sized fan on the back of the case to circulate air

Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

Why the hell do you suppose IBM is helping to Pimp linux?!

Linux is a perfectly capable OS, just as usefull to mom and dad as any other OS (once it's setup and runnning)

I am running Linux on a POWER/PowerPC computer right now.

Cell archatecture is a slight departure based on the Power platform. Designed by IBM, manufactured by Sony.

IBM is the kind of hardware virtualization. You think VMWARE is hot? IBM was doing that in the seventies. Right now you can run hundreds of Linux OSes on a single IBM mainframe. The performance loss is minimal. So by combining cells you have the opturtunity to use hardware abstraction to create a very powerfull computer to run a Linux OS on.

Sony has ported Linux to playstation 2 themselves, so they are used to it and know it.

IBM is releasing a Cell workstation that will run Linux so that Sony Playstation 3 developers have something to work on and run tests on.

Linux has been ported to dozens of different archatectures, the defualt 2.6 kernel has been shown to scale to 64 proccessors. It runs on numerous disparent 64bit platforms (Alpha, Sparc, AMD64, Power, Itanium, etc).

Microsoft is still dicking around with porting Windows to AMD64... a platform mostly compatable with x86. (don't give me crap about NT running on Alpha. It ran on 32bit version, and there was a early beta of W2k that ran 64bit native, but the Win32 API and everything else you use on your computer is and always has been x86-only)

you have TONS of software aviable for Linux. All of it is relatively easily portable.

Linux is very commonly found in high end 3d movie making (Shriek, Lord of the Rings, any newer 3d movie was rendered and partially modeled in Linux using custom and high-end tools.)

It's a proven, seasoned platform. Easily portable, cheap, stable, and it exists.

THAT'S why IBM is pushing Linux so much! They are tired of having their decisions dictated by the limitation of Windows and they are doing something about it.

Cell + Linux, a dream made in heaven. (well geek heaven)

If it is true about Cell being so powerfull at graphical work, then within a few years it would make Linux and Cell the dominate platform for 3d work, and eventually games.

Screw Microsoft, Windows is holding potentially holding everyone back.

Re:Microsoft isn't supporting it? Who Cares?

[fyngyrz]

don't give me crap about NT running on Alpha. It ran on 32bit version, and there was a early beta of W2k that ran 64bit native, but the Win32 API and everything else you use on your computer is and always has been x86-only

It's not crap; we produced release versions of our graphics software for Windows on x86, PowerPC, MIPS and Alpha at one point. Shipped some, too. We had machines for all four architectures (still have them, in fact, though the Alpha and PowerPC's are mothballed), development tools, and working Windows OS's on all of them, and they all ran Windows NT, approximately the same version. Perfect, definitely not -- but Windows under x86 isn't perfect either. It worked well, certainly no worse than the x86 versions. We still use one of the MIPS machines as a backup file server. It refuses to die.

Now, I'm no fan of Windows, but if you think MS couldn't port Windows to another architecture beyond x86, you're only fooling yourself. They can any time they want to, they have already, three times that I know of for certain, not counting whatever credit you want to give Windows CE ports, if any, and there you have it. For all I know there may have been ports to 68k archtectures... I wouldn't be in the least bit surprised.

You have to consider that MS has more money than anyone, and if they decide to go this route, there is no reason to think they cannot do it. I doubt there is any market force, including Sony and the largest governments in the world, that could put a serious roadblock in front of them in this arena.

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

OMFG.

And why the hell would the marketplace give a damn about a MS OS when you can have complete control of your product with open source unix implementations.

Get a clue guy.

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

Get over it, Troll. Linux is good, but not for everyone.

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

Because the actors in the marketplace use tools developed for M$ Window$?

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

Microsoft is still dicking around with porting Windows to AMD64... a platform mostly compatable with x86. (don't give me crap about NT running on Alpha. It ran on 32bit version, and there was a early beta of W2k that ran 64bit native, but the Win32 API and everything else you use on your computer is and always has been x86-only)

There are two operating systems Microsoft have developed called Windows. DOS/Windows, the original one, was based on an x86 clone of CP/M that Microsoft bought. The first version, "Windows 1.0", was released in 1985. The last version, called "Windows Me", was released in 2000, IIRC. This OS was always x86-only, originally ran on archaic CPUs without memory protection and never supported full protected memory, symmetric multiprocessing or other (now) basic OS features.

The second OS developed by Microsoft that's marketed as Windows is Windows NT (now just called "Windows"). It was started in 1988, and never had any relation to DOS/Windows, except insofar as it can (to some extent) emulate it for compatibility reasons (including an x86 emulator on hardware that can't natively execute x86 code). Windows NT was developed on the MIPS platform, not the x86. The original plan had been to use the Intel i860 (an LIW architecture completely different from the x86) as the development platform, but the i860 hardware never met its promise, so MIPS was chosen instead.

The first version of Windows NT was released in 1993, and called "Windows NT 3.1" (3.1 was used for marketing reasons, since that was the latest version of DOS/Windows at the time). Like UNIX, it was mostly written in C, with assembly at the low level to handle hardware dependencies. At its release, Windows NT 3.1 ran on 32-bit MIPS (the development platform) and 32-bit x86 (the first port).

The second version of Windows NT (3.5) was released in 1994, and planned to add 64-bit Alpha (in a semi-crippled, 32-bit mode) and 32-bit PowerPC. However, IBM and Motorola ran into problems with the hardware (in part because of ongoing disagreements with Apple, who wanted to use their own, proprietary platform), so Windows NT 3.5 only added Alpha support. In 1995, after IBM and Motorola had managed to (mostly) sort out their problems (but with Apple declining to follow the IBM/Motorola PReP standard), the PowerPC port of Windows NT was completed, and released as version 3.51. At this point, the OS ran on MIPS, x86, Alpha and PowerPC.

In 1996, the user interface of Windows NT was upgraded to match the user interface of the popular 4.0 release of DOS/Windows (called Windows 95). Windows NT 4.0, which copied the user interface of DOS/Windows 4.0, ran on MIPS, x86, Alpha and PowerPC.

By the late 1990s, as Microsoft continued work on version 5.0 of Windows NT, the market had lost confidence in non-x86 systems for general-purpose PCs (apart from Apple Macs, which didn't follow the PReP standard, so couldn't run OSes ported to it, like AIX and Windows NT). As a result, Microsoft and the vendors of MIPS and PowerPC workstations agreed to cease development and marketing of NT 5.0 for those platforms. Windows NT 5.0 continued to be developed for the x86 and DEC Alpha architectures, into the beta releases.

DEC (which was taken over by Compaq) had continued to have hope for the Alpha as a general-purpose alternative to the x86, but financial difficulties led to the project being abandoned towards the end of the developent cycle for Windows NT 5.0 (marketed as "Windows 2000"). As a result, Windows NT 5.0, completed at the end of 1999, was the first version of NT that only ran on one platform (the x86).

A port of Windows NT 5.0 to the 64-bit Intel Itanium, including 64-bit versions of the Windows APIs (unlike the earlier Alpha port), was released in 2001, but only to select customers.

Windows NT 5.1 (marketed as "Windows XP) was also released in 2001, and again only ran on the x86, apart from another 64-bit limited release for Itanium (in 2002, IIRC).

Windows NT 5.2 (marketed as "Windows Se

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

In the general market for PCs and similar devices, they would give a damn for the same reason they do now: except for tiny minorities, people who buy desktop, and especially laptop, computers want to run Windows. Full Stop. Even in the server market, where UNIX and Linux are strong, Windows is the most popular OS (with over 50% market share last time I checked), and continues to gain market share (so does Linux; the rest are losing).

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

I used NT on ChRP PPC hardware in 1996. It was never at functional partity with the x86 version, and was much less stable.

To be honest, NT on PPC was never over beta quality, and that is being generous. Many applications were missing or unavailable.

Soon afterwards, MIPS and PPC NT versions were killed. Then, eventually, Alpha.

Thus, NT went back to single-arch. Where it remains to this day....

Look at the next attempt by Microsoft to go to different arches: the ia64 debacle for windows server 2000. Again, same story: non-x86 versions suck ass, have stability issues, are chronically late, missing functionality of the x86 versions, etc.

There are real issues with Microsoft and portability, just based on historical experience. People should stop giving them a free pass on this bullshit, and start calling them on it.

It turns out they may not have the tools technology (compilers, linkers, runtime libs) to pull this off for real.

People seriously underestimate the superiority of GNU tools in this regard.

Re:Microsoft isn't supporting it? Who Cares?

[Anonymous Coward]

i would just like the chance to correct this

"By the late 1990s, as Microsoft continued work on version 5.0 of Windows NT, the market had lost confidence in non-x86 systems for general-purpose PCs (apart from Apple Macs, which didn't follow the PReP standard, so couldn't run OSes ported to it, like AIX and Windows NT). "

This is wrong, in the fact there were apple systems did actually ship with aix and not an os from apple. However this were there servers. The two servers that ran aix from apple are the network server 500/700.

Re:Microsoft isn't supporting it? Who Cares?

[Leto-II]

I would say an Apple system running AIX is just that, an Apple system running AIX. Not an Apple Macintosh, which is what the original poster was talking about.

Re:Microsoft isn't supporting it? Who Cares?

[fyngyrz]

PPC NT was extremely functional, and was a very stable applications development platform. We ported a huge application -- something considerably more powerful and functional than Photoshop, though quite different in many respects -- without any real trouble. Certainly no more than was posed by going between Windows 95 and NT.

You can say anything you want, but the fact is, the platform was there and just about as stable as any other NT.

Your remark about applications not being available is certainly on-target, but that doesn't have anything in particular to do with MS's ability to put the OS on a particular hardware platform -- cell, in this case.

The real issue here is that Alpha, MIPS, and definitely PPC were not accepted into the marketplace in any kind of reasonable numbers, and MS made a (most likely perfectly reasonable) decision not to pursue OS development further. Early adopters -- and early supporters, like us -- took a risk and we got burned for taking that risk. But that doesn't mean that there cannot be a situation where a port would take off.

If cell lives up to the hype as typified in the FA, then MS may well decide to indulge themselves in another round of OS building. If they do, it will be huge news, and give the processor a cachet that all the Linux development in the world cannot. Also, if the cell lives up to the hype, we're talking about a whole different level of performance. Some applications -- like ours -- can use every bit of performance that can be provided. Ray tracing is a good example here. Our software, among other things, has an integral ray tracer. No matter how powerful cell is, it is not powerful enough to make non-trivial ray tracing a "no wait" operation. But if it is 100x as powerful as the top of the line x86, it'll sure make a noticable difference. The same thing applies to high resolution (cinema and to some extent HDTV, I mean, not standard video), special effects processing of streams will be much faster too. And that will sell quite a few cell-based machines. If every non-trivial application that is ported shows a considerable speed-up, developers will be motivated to port, because users want performance. What do you think Intel has been making money off all of these years? The user's desire for more performance, that's what. It's really that simple.

I'm certainly not saying that porting is trivial. But I am saying they have done it in the past, and they can do it again in the future. If you want to continue to argue the point, by all means feel free to have the last word -- but you're not going to get me to agree with you. I've seen MS work; I've been deeply involved on three of the ports they've done; and I know they can do it. Furthermore, if the platform appears to be more viable than the older RISC machines, I think they'll almost certainly stick to it longer and work harder on it as well. If there's profit to be made, MS will be looking hard at the situation. Count on it. Microsoft is annoying as hell, but it sure isn't blind or stupid as a corporate entity.

Re:Microsoft isn't supporting it? Who Cares?

[Tony]

The first version of Windows NT was released in 1993, and called "Windows NT 3.1" (3.1 was used for marketing reasons, since that was the latest version of DOS/Windows at the time).

It would have made more sense to distinguish NT from regular MS-Windows by using a new version number (such as MS-Windows NT 4.0), rather than confuse the issue with the *same* version number.

There was another reason for the "Windows 3.1" moniker-- Microsoft had a license agreement with Novell, in which Microsoft was able to include Netware drivers in all 3.x versions of MS-Windows. By labelling the shipping product as a 3.x version of MS-Windows, Microsoft was able to avoid re-negotiating a license.

Since MS-Windows NT was aimed squarely at Novell and Netware, the dominant PC-based networking product at the time, Novell would not have negotiated such favorable terms.

Microsoft maintained the 3.x versioning of MS-Windows NT until Netware was no longer dominant.

Re:Microsoft isn't supporting it? Who Cares?

[Bri3D]

co-ordination to ensure drivers are available, all the popular x86 software runs in the x86 emulation mode without any problems

Um...there is no x86 emulation mode. The AMD64 is x86 with 64 bit extensions. All x86 software runs EXACTLY the same with no emulation. Drivers are the main sticking point(ATi still has sub-optimal performance for no appearant reason).

Re:Microsoft isn't supporting it? Who Cares?

[karlm]

Um...there is no x86 emulation mode. The AMD64 is x86 with 64 bit extensions. All x86 software runs EXACTLY the same with no emulation.

Umm... are you saying that the AMD64 64-bit mode instructions are a superset of the i586 instructions? This is not the case and for this reason there is a seperate 32-bit compatibility mode for running 32-bit applications under 64-bit kernels. (See citation below.)

I think the poster meant 32-bit compatibility mode, not "emulation mode". The differences between a "legacy emulation mode" and a "legacy compatibility mode" are mostly semantic. In any case, the AMD64 instruction set is not a superset of the i586 instruction set, hence long-mode 32-bit ring3 "compatibility mode", which causes the CPU to run in 64-bit mode in ring0, but 32-bit mode in ring3. The FPU programming model is no-longer stack-oriented in 64-bit mode and I've heard that the BCD arithmetic instructions are gone in 64-bit mode. However, I have no personal in-depth knowledge of the ISA changes. Some instruction encodings have definately been eliminated in 64-bit mode. (See citation below)

See "The AMD64(tm) ISA Value Proposition", especially the first sentence in the last paragraph of page 5, which reads "Additionally, some of the binary instruction encodings have been eliminated to support future instruction set expansion." Also see the bottom of page 9, where it mentions the outdated x87 FP stack is eliminated in favor of a more modern FP register file. Clearly, the AMD64 ISA is not a superset of the i586 ISA.

Compare the multi-level 16-bit/32-bit/64-bit AMD64 compatibility modes (including kludges like the A20 gate) to the POWER/PowerPC ISA, where the original ISA was designed to be 64-bit with a pre-defined subset for 32-bit implementations. It's a shame Intel didn't come up with a completely new architecture and create a software emulator to ease transition, a la Apple's switch from m68k to PPC. At least AMD is finally getting rid of the x87 FP stack (except in compatibility modes).

Maybe some day there will be an x86 chip that starts up in 64-bit mode and allow the boot loader to switch the CPU into legacy modes so OS writers can forget all of that legacy kludge (including using a spare pin on the keyboard controller to determine the functionality of the 21st memory bus address line... necessitating keyboard controller chipset-dependent code to change addressing reverse-compatibility...)

Re:Microsoft isn't supporting it? Who Cares?

[thoromyr]

I won't comment about NT specific things you said, but your versions and timings of DOS/Win are just flat out wrong. DOS and Windows *never* had a version match. That is to say, Windows 3.1 (about the first usable version of Windows) had absolutely *nothing* to do with DOS 3.1 (other than run on top of it). Heck, by that time the "standard" DOS was MS-DOS 3.2 and there were probably more current DOS versions (I never kept complete track of when they were released).

DOS *6* was out in 1993 if I recall correctly. DOS 6.2 was the last version of standalone DOS. After that DOS and Windows *were* wedded: previously Windows ran as a GUI shell on DOS, mainly to provided a program manager. Starting with Win95, the GUI shell became integral and was started by default. However, you could still drop back to the DOS shell (start in DOS).

I don't recall for certain the DOS version number in Win95, though as MS had already used 6 I expect they went to 7. It certainly was *not* "a combination of DOS/WIndows 4". DOS was already past version 4 and the Win95 GUI was not a separate graphical shell in the same sense as previous versions of windows. Win95 was an outgrowth from and merging of DOS 6 and Win 3.

So, you see, the DOS version numbers were *never* tied to any windows version number.

thoromyr

Is it just me?

[morriscat69]

Or does the logical extension of this chart:

http://www.blachford.info/computer/Cells/Cell_Dist ributed.gif

Make it look a little more like a HAL than a Cell?

Re:Is it just me?

[benjaminchoate]

Someone please correct me if I'm wrong, but wouldn't the latency of having to send information over a network to your pda/printer/whatever and back make it practically useless for 3D games?

Re:Is it just me?

[SB5]

The way current PCs are built than yes, but if the path was short along with the transmission path being able to send large amounts of data fast and/or the information sent to those places wasn't time dependent, then it could be possible.

Re:Is it just me?

[ajlitt]

Yes. What makes most modern supercomputers based around a pile of off-the-shelf processors so super is the CPU interconnect system (and the I/O system, but that's another story).

Re:Is it just me?

[Mycroft_VIII]

I suppose if you could send 'guestimates' a few frames ahead to several remote cpus for some degree of pre-processing then just grab whichever result is correct at the right time might work, kinda like branch prediction in modern cpus though, if your guesses aren't good enough you waste the extra cpus AND any overhead involved in your guesswork scheme.
Of course to benifit your predictive shceme is going to have to be good enough to pay for it's own overhead.
And of course your could offload less time sensitive things and do more time dependand things locally as another poster suggests.

Mycroft

Re:What's that? Microsoft isn't supporting it?

[djupedal]

Maybe you (and others) haven't noticed, but the desktop PC is a deer in the headlights. Game machines will take over before you can say 'service contract', with networked apps and entertainment ingrained.

IBM knows this....Sony knows this....Apple knows this. Microsoft knows it as well, thus the lack of steam behind Longhorn.

Intel? ...those boys are cashing in stock options and selling die to Taiwan as we speak. Last one out turn off the lava lamp.

Re:What's that? Microsoft isn't supporting it?

[gl4ss]

it wouldn't die, just have a different use.

the article makes so outrageous claims of cell's powers that it makes LITTLE difference if office apps run on it or not. you don't need office apps on your supercomputer.

Re:What's that? Microsoft isn't supporting it?

[darthdrinker]

Indeed, sounds to me like Sony's marketing behemoth is getting into top gears promoting cell in any way possible. Although this might not be directly connected to Sony. Wild claims and theorecal performance papers have been wrong in the past when it came to yet another product with mind blowing specs(Crusoe anyone).

Cells everywhere!

[mrgsd]

The full specifications haven't been given out yet but some details [Specs] are out there:

* 4.6 GHz
* 1.3v
* 85 Celcius operation with heat sink

In toasters.. ovens..

Re:Cells everywhere!

[chendo]

Hell yeah. How awesome would it be to say, "Dude, my toaster will smoke your box at playing HL2!"
I, for one, will fear a Beowulf clusters of Cell-toasters.

Re:Cells everywhere!

[Anonymous Coward]

I think the 85C operation relates to allowed ambient. It is a function of the heat generated by the silicon over the thermal resistance of the package. If the power dissipation or the Thermal resistance increases, then the ambient tempurature required to keep the silicon below its failure point goes down. The higher the ambient operating tempurature of the part, the better. It indicates either a) low power dissipation or b) good thermal packaging design. (probably a little of both....)

Is that really that revolutionary?

[drgonzo59]

The idea of having many processing units in a personal workstation is not new. They thought that Moore's law was going to fail years ago and predicted that by now we would all have massively parallel machines at home on our desks. Well it turned out that Moore's law didn't fail and most importantly that many of the software algorithms are not easily parallelizable. So what if I can have 100 cells at home in my workstation. I could run SETI, weather or some other kind of simulation but I couldn't really play my video games much faster or have a more responsive user interface if I ever install Longhorn. I just can't think of too many programs run on home user's machines that would benefit form a parallel architecture.

Now if the can be made very fast and have only a few (2-8) coupled together...well,as it was said, that is what a nice Opteron machine does anyway nowadays.

Re:Is that really that revolutionary?

[rbanffy]

Well it turned out that Moore's law didn't fail and most importantly that many of the software algorithms are not easily parallelizable

Well... It turns out that most videogame and simulation code is easily parallelizable and that our desktop computers are running a couple dozen programs at the same time (i count 8 active ones right now and I just woke up).

I would buy a Cell workstation running Linux, even knowing it would never run Windows, just because I like interesting hardware. Given a cheap enough hardware and enough software resources to round out existing software, how long do you think it would take for someone to build The Killer Home Computer

And, if Apple wants it, just think of the Mac Mini they could build with it.

Re:Is that really that revolutionary?

[Anonymous Coward]

"It turns out that most videogame and simulation code is easily parallelizable"

Shhh!!!

Let's let all the fucking x86 clowns keep repeating the "hard to program" BS...

Re:Is that really that revolutionary?

[rbanffy]

Well... The lack of registers sure makes the x86 a bit harder to program than more modern architectures. It is also due to Microsoft's support (or lack of it) for true decent multitasking computing until just about er... never, that the x86 has been optimized well beyond all reasonable efforts for single-tasking compute-heavy applications. There was no point on a high throughput x86 system until BSD and Linux came around. If you wanted high performance and was willing to pay branded Unix prices, you would go SPARC, POWER, PA or Alpha.

Re:Is that really that revolutionary?

[Anonymous Coward]

Windows NT and UNIX System V were very scalable (for the hardware) SMP systems on x86 when Linux and BSD lacked any SMP support al all. Where do you think all the predictions of NT killing UNIX came from? Of course, NT and UNIX System V for x86 were no match for high-end RISC systems running UNIX (not Linux), but that was simple the reality of the PC hardware (which is a different story today).

Windows NT (Windows 2003) and UNIX still scale better on the x86 than Linux, and certainly better than BSD, which is only beginning to become decent on systems in the 2-4 CPU range.

Linux and BSD are great systems, and I've long been a user/developer on those two platforms, but to ignore the reality that commercial systems like Windows NT, UNIX System V and Solaris have always had better SMP implementations (and had them first), and have always scaled better on x86, is simply a combination of ignorance and rabid fanaticism.

Re:Is that really that revolutionary?

[flaming-opus]

No it's not at all revolutionary, but you stumbled upon one of the rare bits of insight in the world of computers: NOTHING NEW EVER EVER EVER HAPPENS IN COMPUTERS!

We figured out the basics for how to build a computer by about 1970, and since then there have really only been about three discoveries per decade. Most of those aren't even technology discoveries, but economic or marketing break-throughs. Most of computer engineering is taking old ideas, and incrementing them, or packaging them together in different ratios, or in new combinations.

Is the cell architecture new? No! It's 8 DSP's chained to a special I/O port on a powerpc processor. Each has a dedicated block of sram, and can mark dirty pages in the powerpc's main memory. Actually they probably can't, I bet the master processor does that. Will this thing be really fast for decoding mpeg streams, yeah probably it will. Will it be fast for general purpose supercomputing? - Absolutely not - supercomputing is memory limited, not ALU limited. Will it revolutionize servers or desktops? - absolutely not, they are limited by branches, and unstructured memory access, not by computation ability.

The cell is a really powerful brute-force solution to a problem that not many people have. It's probably a great fit for ps3, and not for a lot else. If it were, you'd see IBM promoting it in other product lines. They aren't, so it's obviously not that usefull outside of embeded dsp-style applications.

Re:Is that really that revolutionary?

[rbanffy]

My point was (and still is) that there wasn't a high performance OS for the x86 in the PC price-range available. It's true Windows NT scaled to lots of CPUs, but it does so only on extremely costly and exotic hardware and the license costs were and still are outrageously high. The same was true about UNIX System V and other branded Unixes.

The limitations of software had always limited the development of mainstream hardware and multiprocessors are no exception. Because of the high price of OSs that could take advantage of CPU level parallelism, mainstream processor vendors chose to optimize their hardware for single-threaded performance. Were there a quad-486 on a chip instead of a Pentium or an 8-way unit instead of Pentium Pro, we would have very different software today. Of course, they would flop because Windows 3, 95 and 98 didn't support multiprocessors (MacOS Classic doesn't have a much better story to tell about it either) and NT was way too expensive (and perceived as slow and more complex). Unixes of the time wouldn't qualify as mainstream either. And so, nobody ever tried to put more than one core on a mainstream consumer-level CPU.

Re:Is that really that revolutionary?

[Anonymous Coward]

In most cases, it's much harder to write an application in such a way that it will scale to multiple CPUs. Except for special cases, the payoff is usually not worth the effort, which is why most applications have been, and continue to be optimised for single-CPU systems. Where multiple CPUs are most useful is in running multiple applications.

Compiler technology

[sifi]

One question which was not addressed fully in the article was how do you compile/test programs for this thing.

The potential of parallel architectures has never been in doubt since the early days of the Cray monsters - but how to compile code to use all the features efficiently has.

I don't believe that we see the full advantage of these types of architecture exploited without some similar break-through in software tools.

Mind you the hardware rocks...

Re:Compiler technology

[mr_jrt]

Program in a language that is referentially transparent.

...once you can assume that any function is able to be concurrently executed all you have to solve is the communication between processors/storage. The latency of current networking technologies makes this unpractical for general tasks, but this is less of a problem with a low-latency internal bus.

Time to drag those Haskell textbooks out of the closet and dust them off. ;)

Re:Compiler technology

[faragon]

Sony's Playstation 2 had a similar archiecture (4 CPUs: R3000, R5900, 2xVU), paralelism is achieved by specializing each CPU for a concrete task. Then, you have no need for thinking how to parallelize one task, as no task are parallelizable, but to find a task set distributabe to the CPUs (thinking that each CPU has his "virtues" and his "defects", being ideal for some tasks and inneficient for other, etc.).

Re:Compiler technology

[sifi]

...and the PS2 had (has) terrible software tools, developers *hate* programming for it compared with other consoles.

There is no way you can write high performance software for the PS2 without knowing about the internal architecture specifics.

Re:Compiler technology

[faragon]

Well, I'm used to program the PS2 in my spare time, with such terrible tools (gcc for the R5900, vcl for the VU's), enjoying a lot with it.

There is no way you can write high performance software for the PS2 without knowing about the internal architecture specifics.
->Shouldn't that be the natural situation? If you want to optimize a solution for a given system, you should know a bit that system. One of the "bigger" problems of the PS2 has is that the main CPU has a few cache memory, forcing developers to avoid bus stall situations, etc. but not a computer architecture issue.

For finishing, if you're not teached about one system, learn first its essentials, before blaming about "how hard to program it is".

Re:Compiler technology

[Vacindak]

Just make them all write their code in Scheme/Lisp.

Re:Compiler technology

[Vacindak]

And now I'm curious whether people will mod that Funny or Insightful.

Re:Compiler technology

[Bloem]

Better yet, use occam !
http://wotug.kent.ac.uk/parallel/www/occam/occam.h tml

Re:Compiler technology

[Anonymous Coward]

Of course! Don't you remember how you are supposed to be able to plug in your toaster so that your PS3 can offload processing to it?

Re:Compiler technology

[GileadGreene]

The breakthroughs that you want were invented years ago. You might want to look into a language called occam - see occam-pi for a recent implementation. Occam is based on Hoare's CSP concurrency model, which makes dealing with concurrency easy. Similar ideas were implemented in Plan 9's Alef language, and also in Limbo on the Inferno OS.

Re:Compiler technology

[kill-1]

But what if you have n referentially transparent functions and your main program looks like

main(x) { return f1(f2(f3(...fn(x)...))); }

Re:Compiler technology

[mr_jrt]

Then the program can be solved at compile time and all it spits out is a constant value :)

Re:Compiler technology

[mr_jrt]

D'oh. My apoligies...just noticed the argument to main. In that case you are indeed limited to the capibilities of a single processing unit. I would have to say though, thats the worse case scenario, and the same problem can be said of even efficient algorithms, (eg. binary trees which are not rebalanced frequently, etc.)

Serial and Parallel

[Anonymous Coward]

No matter how well a processor or group of processors can run tasks concurrently it will always come down to the fact that most tasks are serial in nature and will not scale to a concurrent processing architecture. Aside from this developing multi-threaded software is extremely difficult and is rife with problems. Just ask any developer about the hardest problem to find/debug. It is pain incranate and some MT bugs can take 5+ days to find. People design serially, because a lot of tasks are essentially serial in nature, and until this design paradigm gets a major shift and we design parallel only software [LOL] then cell has no future.

Re:Serial and Parallel

[Icarus1919]

If you had read TFA you would have read that cell performs even better with serial processing than it does with parallel processing.

Re:Serial and Parallel

[milohanrahan]

That's not really the point, though, is it, and certainly not the one he was making.

Re:Serial and Parallel

[Anonymous Coward]

This is 10 years out of date.

Developing MT software is rife with problems if you do it from first principles using threads and semaphores and make no attempt to design a system that is robust. In other words, a beginner at MT can get very very stuck but professionals have been writing reliable MT programs with no more bugs than normal for at least 5 years.

For instance, MT-aware servers all use the worker-thread model. It's not the full generality of what you can do with threads and semaphores but it works and it's reliable and easy to debug each job since they are independent.

I just know that half a dozen game developers are going to try to write a program that has 20 different threads all in different thread functions and all interlocking manually. It's going to take more than 5 days to find the bugs in that!

But it's nothing to do with MT, it's to do with people's lack of understanding of how to write MT code properly.

Re:Serial and Parallel

[diamondsw]

However, most graphics and AI problems are highly parallel in nature, which is exactly where video games are going.

Re:Serial and Parallel

[GileadGreene]

Multithreaded programming is only hard if you use poor tools. If you are trying to write multithreaded programs using ancient concepts of concurrency like mutexes and semaphores then yeah, it'll be hard. If you write software using more modern concepts like CSP, and languages or libraries like occam-pi or JCSP, then multithreading can become extremely easy. I've written reasonably complex multithreaded programs with several hundred "threads" (some nested inside others) running simultaneously and interacting with each other (not just a limited worker thread model) and faced no more debugging difficulties than I would have writing a sequential program of equal complexity.

Re:What's that? Microsoft isn't supporting it?

[a3atom]

Who cares? Mac OS X and Linux will provide all the applications required. Windows apps will be likely be available under emulation. The Windows market will still dominate but there will be a gradual migration when people realise there are cheaper/better realistic alternatives available at last.

Right at the end of the article

[baker_tony]

"the Cell means a decade from now the technology world is going to look very different."

Wow, and just think how the "technology world" 10 years ago was basically identical. I think they could've used a better phrase here.

Re:Right at the end of the article

[MrRuslan]

Well yes it wasnt that difrent...we had everthing the same way as we do now...x86 based cpu dominating,Windows crashing,Optical storage,etc. only now they are more pumped than they were then.Now if cell kills x86 somehow (I hope so) computers will me alot difrent.

Re:What's that? Microsoft isn't supporting it?

[Ohreally_factor]

Especially now that Sony has decided to be a hardware company once again.

next please

[aixou]

I'm sorry, but Sony can kiss my ass.

This is from the company that said the Playstation 2 would have Toy Story quality graphics, and be able to render FF8 quality FMVs in real time (thus making FMVs no longer required). It was essentially that bullshit hype that killed the Dreamcast... so yeah, now they're at it again.

Maybe I'll be proven wrong, but I doubt their system will be able to do anywhere near what they say it can in practical application.

Re:next please

[Anonymous Coward]

Games on dreamcast looked BETTER renderred tahn on ps2 crap for a while

Crazy Taxi on dreamcast vs same game on PS2

enough said

for REAL hidef gaming try xbox (certain titles)

Re:next please

[binary42]

IMO FF8 is well beyond Toy Story in quality and complexity. I do agree that Toy Story is a bit much... though if you ever watch that movie again you will be suprised by its simplicity compaired to current day films. I would say that PS2 can do amaizing stuff people just keep steping on it because some stupid marketing people opened their mouth. It did unfortunately take then a LONG time to master the system and its design.

Re:next please

[Anonymous Coward]

Christ, I really hope it is the same AC troll writing these moronical posts in answer to every poster daring to question if the Cell will live up to the hype.

Re:next please

[Ziviyr]

And I still have yet to see texture filtering typically used on that system.

Actually, I keep hearing about SIMD units on it, but no real 3D specific hardware on it has come to my attention. As they say, WTF?

Re:next please

[Rinikusu]

Sony can kiss my ass, too. But I'll probably be in the fucking line to buy it when it comes out. See you there?

Re:next please

[Anonymous Coward]

Fucking Google it. Of course it's got real 3D specific hardware. You think they're software rendering 15 million polys per second on a 300MHz CPU?

Re:next please

[Abcd1234]

That, of course, would be the brilliance of Pixar. Because they do cartoon-style animation, they can get away with a reduced level of realism (and thus decreased rendering complexity). This also means people are more... err, satisfied with their work. After all, there's nothing worse than seeing a poorly rendered human (in fact, there was an article on this very topic a while back, though I can't find the link), as opposed to a cutely rendered cartoon fish. Of course, that's not to say they don't do damned impressive stuff... the hair rendering in Monster's Inc. was really breathtaking, and their work in Finding Nemo was pretty impressive, too. But they definitely don't aim for photorealism.

Re:next please

[Jerf]

Maybe this time Sony will see fit to include that really high-tech mipmapping stuff, so their console isn't the King of Sparkle.

(Stupid Sony, I've had my PS/2 for about a year now and I still notice it almost every time I play. Can't believe how unbelievably stupid they were not to include it. That one change, which by computer graphics standards is dirt cheap, would have massively improved its graphics. Anti-aliasing, on the other hand, is expensive to do right, so while I expect it on this next generation, at least while running in NTSC or PAL, I wouldn't have expected it in the PS/2 era. Though some managed, I think....)

After that, I don't trust them any farther than I can throw them. The PS/2's graphics subsystem wasn't an Eighth Wonder of the World, it was an incompetent disgrace. Fortunately most of their fanboys are so stuck up the ass with Sony that it took them years to notice, instead of it jumping out at them in 5 seconds.

I have it for the game selection, and I like the games, I like the controller, I like the case, etc... but the graphics are far, far worse than what they should have been. You have to reach back for years and years to find anything else that didn't do mipmapping.

(I've also played the Dreamcast some more lately. It definately pumps out fewer polygons, but equally definately, they are higher quality polygons, and the fact that the Dreamcast clearly has mip-mapping is no small part of that. The PS/2 was a step forward in some ways, but a big step back from the DC in others.)

Re:next please

[Anonymous Coward]

After all, there's nothing worse than seeing a poorly rendered human (in fact, there was an article on this very topic a while back, though I can't find the link)

http://www.google.com/search?q=uncanny+valley

Re:next please

[protohiro1]

The difference here is a)This article is not a claim, its a poor distillation of a patent app and b) This other company is involved. This list seems to indicate they might know something about fast processors. And possibly more than the average slashdot poster.

Re:next please

[Abcd1234]

Yes, that's the one! Thanks! Into Del.icio.us it goes... :)

Re:next please

[Anonymous Coward]

Yes, you will be proven wrong. No doubt about that.

Re:next please

[Anonymous Coward]

...another sad DC owner bites the dust.

Re:next please

[Anonymous Coward]

Actually, your statement is misleading, as the PS2 does support mip-mapping. What it doesn't support is triangle angle dependant calculation of mip map level. It computes mip map level based on homogeneous W (or it can be set to a constant). This does produce a lot of aliasing, but nowhere near as bad as what you would see if there were no mip-mapping at all.
In fact, the PS2 has trilinear filtering (at an extra fill rate cost).

err

[aptenergy]

There's a lot of weird stuff in the conclusions part where he compares the Cell against x86 and PowerPC 970 chips. Look at the Apple blurb for example. He's trying to paint this rosy picture about how Apple can sell a bunch of cloned Macs with Cell processors... that's just foolish. Apple has stuck with the PowerPC architecture for a long, long time now; there's no way they would rewrite OS X (and force everyone to rewrite their apps) just to make a bunch of cheap clones. Not just that, but Apple is now well known for outstanding hardware design. No way they're going to license anything to a cheap, beige-box manufacturer. So from my standpoint, his last page was a big load of BS...

Re:err

[Anonymous Coward]

If you bothered to read it.

Cell IS PowerPC. Or escentially PowerPC in nature.

The PowerPC proccessor is something that originally came from IBM, remember?

Basicly you have one Power970 proccessor surrounded by numerious special purpose vector proccessors.

They strip away all the abstraction (The ISA, if you know what that means) and replace it with a special purpose OS devoted to handling the hardware.

IBM has large amounts of virtualization software technology. Like what VMware is, but they've been doing it for over 20 years in their mainframes.

They have servers were you can run dozens of operating systems in one machine with almost no slow down. No overhead, no loss of performance like used in VMware or harware emulation. This is hardware specificly designed for this purpose.

OS X or Linux can run in a proccessing partitoin on a Cell-based computer with little effort porting it from PowerPC. IBM already does this with Linux on mainframes and it wouldn't be difficult to do this with OS X.

Basicly the OS will SEE a PowerPC computer, but it will be cell and the underlyning software "controller" will take care of all the details of sending proccessing power to the various special purpose cpu's in the 'cell'

Basicly you can have something like 8 PowerPC "cells", each cell made up of 5 or more cores. One core will be almost like the Power970 proccessor (apple g5), and will seem that way to the hosted Operating systems and the rest will be special purpose proccessors.

That's why they can run so fast, they are special purpose, no memory controller like in AMD64, no x86 ISA constrants, no SSE, no nothing. Just Vector proccessors.

(ISA = hardware compatablity abstraction. This isn't ISA slots or ISA bus, this is the layer of hardware that allows modern computers to do things like still run DOS, even though DOS was designed to run on 386 cpus. With the Cell this gets replaced with software, the amount of overhead this causes is not that much considuring we are dealing with 3-4ghz cpus.)

Re:err

[Anonymous Coward]

Unless OS X is written in assembly language they won't have to rewrite it. They would recompile it and fixup the parts that break. If the cell hardware is as powerful as the article says they can emulate Apple hardware. Forgetaboutit.

Re:err

[ant_slayer]

I'm thinking that you may not have read the article. The "proposed" architecture uses what is essentially a G6 (same architecture as the G4/G5) as its "Processing Unit". Adding the vector processing units doesn't change that. OS X should basically run on the G6 and offload streamed processing threads to the vector units.

As the author points out, multimedia operations are already passed off to vector processing units in SSE and Altiva style multimedia instruction pathways with today's architectures. This occurs both on the main CPUs and within the GPUs from Nvidia and ATI.

The new twist is that IBM, Toshiba, and Sony are moving this power into the CPU and taking out a bunch of the old architectural assumptions that have hindered our aging x86 architecture. By pulling memory in larger blocks, addressing it differently, killing the cache and moving its power into private memories for the vector processing units, there's a lot going on to make this move.

My gut feel is that the response to the questions you raise as well as many of the afore-mentioned beefs about how "hard" it is to program for a parallel environment is that it won't be that tough. We already have a multiplicity of processing streams running on our machines -- it's just that one is on the CPU, one is on the GPU, one is in the sound card, etc.

The impact for the games developer or the flashy-GUI-front-end developer for the OS, or the multimedia-content-generation-software developer could conceivably be small. Regular code is serialized in the main processing unit, while well-written libraries spool vectorizable and compartmentalizable tasks off to the vector CPUs.

How many of us actually use our multi-gigahertz CPUs to their potential anyway? The big, demanding stuff we do with our PCs is *already* threaded off to other processors, abstracted in our software world so that we aren't controlling those individual threads of execution. The whole point of the GPU is that we don't have to micromanage the processing of it -- I don't think it's impossible for this technology to eventually filter into many of the computing markets.

It's not a panacea, just a new way of looking at things. Will we be seeing 150 GFLOPS as the author predicts? I *really* doubt it. But we may see some really powerful additions to the flashy and computationally intensive components of our computing experience.

Remember -- vector coprocessors already exist, just in a handicapped form. Cell is just the faith-healer for vector coprocessing units ;-).

-Ant Slayer-

Re:err

[Andy_R]

That's because the artictecture is (supposedly) a PowerPC core plus 8 APUs. If this is true, the necessary re-writes would be no more complex than they were the last time Apple bolted on some vector bits to the PowerPC, with Alti-Vec (just a hell of a lot more powerful).

Re:err

[Anonymous Coward]

Gah. Step to the back of class. The Cell uses a POWER core, so OS X can just run straight on it. No one has to port anything.

They'll build it themselves though. You've got that right.

Re:err

[EvilTwinSkippy]

In addition to the other comments posted, you should know that OS X is derived from a lot of work done for the NeXT operating system, which ran on x86. The Kernel and base operating system for OS X is actually NetBSD, arguably the most portable unix out there.

And Apple is all about locking down designs with proprietary hardware. Yes, it uses PCI ports, Power PC procesors, and off-the-shelf ATA and Scsi drives. But sewing all that together is a chipset of their own design.

How could they possibly do this cheap?

[brett42]

I'm willing to believe that a 4.6 GHz chip with 8 ALUs and high bandwidth memory would be fast, but even in bulk, there's no way they can afford to put 4 of them in a sub-$500 game console.

I've been reading PR about the Cell for years, and nothing I've ever read has seemed even remotely plausible. Is there any objective information that even comes close to substantiating any of these claims?

Re:How could they possibly do this cheap?

[Anonymous Coward]

Just an endless stream of dumshits and fanboys everytime a console story gets posted. Sigh.

Here's a hint Einstein:

Why do you think Cell is able to run a such a high clock speed? It is a step back in complexity compared to the tangled mess of current desktop/workstation chips. And guess what happens to manufacturing costs when you simplify things...

Why don't you stop reading 'PR' and 'unsubstantiated claims' and fucking read the patents and other hard info about Cell that has been out there for months retard.

Re:How could they possibly do this cheap?

[Anonymous Coward]

Wow, uhh, relax there mate.

Re:How could they possibly do this cheap?

[Anonymous Coward]

Why do you think Cell is able to run a such a high clock speed? It is a step back in complexity compared to the tangled mess of current desktop/workstation chips. And guess what happens to manufacturing costs when you simplify things...

Perhaps, but that also means cost of developing software for it will increase since memory managemant etc will have to be done explicitly for each application, which in the end might mean the total performance is lower than for a workstation chip.

You are instantly buying into the PR performance numbers announced about a chip that is at least a year away, and you are calling other people fanboy. Whatever.

Re:How could they possibly do this cheap?

[Anonymous Coward]

Bzzt!

Give you one guess what I do for a living...

Re:How could they possibly do this cheap?

[Dutch_Cap]

"I'm willing to believe that a 4.6 GHz chip with 8 ALUs and high bandwidth memory would be fast, but even in bulk, there's no way they can afford to put 4 of them in a sub-$500 game console."

Perhaps this could, at least partly, be explained by the fact that most console manufacturers actually make a loss on console sales, but make up for this by selling lots of games.

Re:How could they possibly do this cheap?

[Anonymous Coward]

"the fact that most console manufacturers actually make a loss on console sales"

Why can't we all live in a universe where people spontaneously died whenever they posted bit of wisdom?

Re:How could they possibly do this cheap?

[Anonymous Coward]

They can't.
As always, CPU costs relate to the size of the die, and the number of defects per wafer - in otherwords yield /real-estate space. Heat generated per square inch is about as high as one wants. Logic=transistors=heat. Quicker = more heat, and a lower manufacturing yield.
Memory speed bogs things down - rosy numbers means that the instruction 'fetch' is not happening. Now if the core was water cooled, then yes, options do open up a little. Programs that run entirely in cache, are quickish. Now you can piggy-back several wafers, glued in with interconnects - good for game, graphics, and encryption/seti like tasks. As pointed out, n-way DMA access shoots security and memory protection to bits. A Cell's advantage, is that the hardware/DRM model should endure a couple of years at least.

Re:How could they possibly do this cheap?

[duffahtolla]

there's no way they can afford to put 4 of them in a sub-$500 game console.

I'm thinking there's no way they can afford not to.

MS wants to take out sony as the leader in the game market. Their tactic is to take cheap PC hardware, snap up hot games and make them exclusive by buying out the game companies. Throw a lot of advertising on it and vio'la. Anyone can see the writing on the wall.

Sony has to come out with something that will smoke whatever the XBox2 is going to be. Not just your regular smoke, but the rip of its head and crap down its throat kind. Massive parallelism is probably the cheapest way to do it and the added bonus is that MS can't duplicate the tactic with their cheap x86 hardware.

Make an Xbox2 look like a toy in comparison to the PS3 and the fight will be over (at least for a while).

With MS as a competitor you either crush or be crushed.

Re:How could they possibly do this cheap?

[The Warlock]

Considering that the PS2 was hyped as being an excessively powerful console and ended up being ridiculously underpowered compared to everything except (maybe) the Dreamcast, I really doubt that we're going to see anything groundbreaking here. At least in the CPU tech department. Anyone doing research in PR techniques can learn a lot here, I guess.

Re:How could they possibly do this cheap?

[KirkH]

Eh? MS is leaving x86 behind for the Xbox 2. They're going with some type of PowerPC based chip from IBM, rumored to be multi-core. ATI provides a custom graphics chipset that will not have a PC counterpart.

Sony is going with Cell from IBM and an nVidia graphics chipset. So I don't see a huge difference. My guess is that both consoles will have extremely similar performance and this next generation of consoles will be the most boring ever -- lots of multi-platform games that look identical.

Re:How could they possibly do this cheap?

[mwvdlee]

So just how much would you expect the manufacturing cost to be for one Cell processor?

Re:How could they possibly do this cheap?

[Anonymous Coward]

I think I know... suck dick! You are what you eat, as they say.

Re:How could they possibly do this cheap?

[Anonymous Coward]

Its called the gillette model.
Sell the console for nothing, make their money on the media..
Sony are making a loss on every PSP they sell, keeping in mind they will make their money back, in no time, with sales of the media...

Re:How could they possibly do this cheap?

[flaming-opus]

This is all extremely plausible. remember: these are not 8 full function ALUs like you get in a modern microprocessor. They opperate on 32-bit single-precision floats only. For all we know they are each in-order long pipelines, with static branch prediction, no speculative execution, a limited distributed memory scheme etc.etc.etc.

It's really 8 digital signal processors attached to a powerpc chip. You can do that today, you just run the data over the PCI bus, rather than on a big in-chip bus. IBM is just gonna stuff all that onto a chip, and hope that 65nm process will make if fast/cool/cheap enough to put in a console.

It has always been possible to make very fast parallel architectures if you make them very special-purpose. Cell sounds like it will be very difficult to program, and only a couple of game-engine designers will bother trying. The rest will just license those engines and work on game-play, story-line, and art. Same as before, only faster.

Re:How could they possibly do this cheap?

[orfanotna]

this next generation of consoles will be the most boring ever -- lots of multi-platform games that look identical.

That's probably true. I hope Nintendo has something great up its sleeve.

Re:How could they possibly do this cheap?

[freemacmini]

If sony was smart they would aim this shot right at the PC/OS market. Put together PCs with the cell processor running linux and place them in every retailer with a sharp demo and see MS sweat bricks.

Re:How could they possibly do this cheap?

[gl4ss]

overhyping worked for sony in the past. so they're doing it now again.

nobody can prove them wrong untill ps3 is out and that's their ace.

say what!

[Anonymous Coward]

Clippy needs 2 cells to run!

Re:What's that? Microsoft isn't supporting it?

[japhmi]

I think the guy who wrote this is over-hyping the cell, but I wouldn't be surprised if the cell smoked everything out there for things like video, audio, and 3D graphics.

Sony and Toshiba may end up owning the living room, while Microsoft etc. fight it out for the desktop. Everyone in the PC industry thinks that the living room will make them more money...

Re:Wow, what crap!

[Anonymous Coward]

He's only posting here because he dosen't have 10 friends!

4x4

[courseB]

future computer ads might sound like '4x4' all-wheel drive ones; 'transfers bits from the cells that grip, to the cells that slip'

Re:4x4

[AddressException]

Don't automotive AWD systems transfer power from slipping wheels to non-slipping ones? ;)

STI

[smallguy78]

i didn't understand any of the document, but damn it looks fast

like Transputer?

[Anonymous Coward]

Does anyone remember the Transputer technology? It was also based on some cell computing approach and was also meant to replace the PC architecture one day.

Listen here

[Anonymous Coward]

WHat most slashdotters fail to realise is that a vector processor would blow chiunks if used for general purpose processing and hence you won'e be seeing a desktop based on them anytime soon.

Sure they may have vecotr processing in the form of SSE etc but vector processing is of no use tfor word processing period!

Re:Listen here

[chendo]

If I read the article correctly, the individual cores have the ability to act independantly of each other, hence it can also act as a normal multi-processor box in the normal sense.

Re:Listen here

[Anonymous Coward]

The the main 'core' of the cell is a Power970-alike cpu. That will be adiquate to run a word proccessor at acceptable speeds, even if it was only running at 700-800mhz.

It's all about 3d stuff and massive proccessing power for games, virtual realities, and multimedia.

Re:Listen here

[Anonymous Coward]

who the fuck told you that people don't rip music or dvds or play with their porn pictures or anything else for that matter...

big mouth

Re:Listen here

[Anonymous Coward]

Why aren't games multithreaded?

Re:Listen here

[Nebu]

Sure they may have vecotr processing in the form of SSE etc but vector processing is of no use tfor word processing period!

Word processing doesn't need acceleration. Games do.

Re:What's that? Microsoft isn't supporting it?

[Yoda's Mum]

It only performs like 20 opterons in highly parallelisable tasks. Which excludes almost every task performed on the average PC, with the exception of some gaming graphics tasks (which, incidentally, are performed on specialised GPU's which vastly outperform x86 cores for their tasks anyway). Most of the time, a single cell core will perform pretty much identical to the single Power chip that controls it.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

Also it helps to understand how IBM runs it's mainframes.

You have virtual PCs. These are are programs like VMWARE so that you can run multiple computer OSes.

Well IBM has similar technology, except that while Vmware has been doing it's thing for a while now, IBM has been doing this for almost 30 years. They have gobs of experiance.

So look at the big mainframe. A beast. The CPU isn't all that powerfull but it's main strength is MASSIVE amount of I/O bandwidth. In a PC you have PCI buss that has a maximum bandwidth of 130 or so MBps. A big new IBM manframe has 20GBps worth of bandwidth.

You have the main OS that runs on bare metal. This is called VM. Inside this OS you have things called 'partitions'. These are not like harddrive partitions, but these are proccessing partitions.

In each of these partitions you can have a OS run in them. You can run OS/370 for instance, for legacy batch programs left over from the 80's and early 90's. I am sitting right next to machine that runs a similar setup for JCL programmers.

However you can run Linux on them, too. You can run HUNDREDS of Linux OSes in them, actually.

And the significant thing is that you can do this with almost no overhead. Programs run as fast, or potentially FASTER, then when they are run on "bare metal".

Now look at the Cell archatecture. You have a Power970-style CPU that runs as a controller for all theses special purpose vector proccessors.

You have a VERY high degree of control over system. Unlike x86 there is almost no abstraction...

Take virtual memory (if you think that VM = Swap space, I am talking way over your head) address spaces. You have a abstract memory space, UNTIL it's being used then it's hardcoded and most of the control is done by the host OS...

So this is what I see as the future for the Cell worlstation:

You have the Cell hardware. Ontop of that you have the special IBM OS that controls the hardware. In proccessing partition like you have with Mainframes you have either a Linux or OS X operating system that runs all the user's applications. You can run one OS, or numerious operating systems at the same time.

To the OS X/Linux operating system you will multiple CPUs that it will access. These multiple CPUs will be the 'cells'. To the OS, to the user it will appear to be a run-of-the-mill PowerPC. Just like a IBook or a PowerG5

You will have the ability to render something like similar to the Doom 3 in real time in OpenGL SOFTWARE RENDERING MODE 10x's faster then the fastest PC aviable anytime in the next 5 years.

Hello, folks. This is the speed in which these guys are talking! This isn't oh, Most powerfull PC in the world bullshit talk like from Apple. This shit is the next level. This is star treck shit, this is your-computer-is-now-HAL, type stuff.

Wow

[iamnotacrook]

Why do you post tripe like this?

FYI Crays come (came) with huge quantities of libraries, preprocessors, compilers, schedulers, profilers, all designed to help efficiently use the hardware.

Don't post comments they say nothing else than "I dont understand."

Re:Wow

[Anonymous Coward]

Don't argue with him. I'm sure he is getting his info from an authority on embedded multimedia hardware like John Carmack or Tim Sweeny...

Re:Wow

[sifi]

Please feel free to correct me if I wrong (I'm sure you will!). My comments were relating more to the multiple cells than the internal vector processors on each cell.

Cray type architectures are basically vector parallel processors. In this case it *is* possible to get compilers to pull vector type operations out of tight serial loops (gee this looks like a vector multiply etc.).

Even then a programmer either has to:
a) Use well defined libraries (vector operations etc.)
b) Know something about the hardware to write loops in the 'correct' way for the compiler to figure out what's going on.

A more difficult problem is how to compile for 'loosely coupled' architectures (like the cell) - Even if each individual cell contains vector processors.

Now you have to write what effectively amounts to large amounts of multithreaded code - behaving cooperatively on a system with an unknown number of nodes.

Now you either need:
a) A really intelligent compiler
or
b) A really intelligent programmer

Re:Wow

[ultranova]

Even then a programmer either has to:
> a) Use well defined libraries (vector operations etc.)
b) Know something about the hardware to write loops in the 'correct' way for the compiler to figure out what's going on.

Is there any system where you can get good performance without knowing anything about the hardware ?

Now you have to write what effectively amounts to large amounts of multithreaded code - behaving cooperatively on a system with an unknown number of nodes.

Is this really that hard ? All you have to do is separate the state at the next step from the state at current state, and synchronize so that

1. Wait until all threads reach this step.
2. Figure out for each gameworld object it's state at the next time step. Store it separately from the current state (in extra fields or whatever). No globally visible data gets changed for any object, so there's no need to synchronize inside this step.
3. Wait tuntil all threads reach this step.
4. For each gameworld object, update the object's current state. Each object is completely self-contained in this step, so they can be updated in parallel.
5. Return to step 1.

The main problem I see with this approach would be the increased memory consumption for storing the extra data (future state), which might cause serious problems with consoles.

The main advantages are simplicity, no need to have per-object locks, and no possibility of deadlocks (unless some thread gets stuck to infinite loop, in which case you're screwed anyway). Also, it should scale to n threads, where n is the amount of objects in the gameworld.

Re:Wow

[stevesliva]

One apparent guess made by the author is that each cell chip will have a single traditional GPU core and 8 of the APU cores. Sounds like compilers (XLC for PPC?) could ignore the APUs necessary.

Re:Wow

[stevesliva]

s/GPU/PPC/

Re:Wow

[fizban]

Um, what if objects interract with each other? You can't simple figure out each and every object's next state in parallel, because each object's state may be dependent on the next or current state of objects it interracts with. Therefore, you *do* have to synchronize during step 2 for any objects that you know have dependencies on other objects right now, or you know *could* have dependencies on other objects in the future.

Re:Wow

[ultranova]

Um, what if objects interract with each other?

Of course they do; if they didn't, the gameworld wouldn't be a very interesting place ;).

You can't simple figure out each and every object's next state in parallel, because each object's state may be dependent on the next or current state of objects it interracts with.

This is why we separate the current state of each object from it's future state. The current state is invariant through step 2.

Do you understand ? All objects (the parts of them that can be read by other objects) are read-only during step 2. No alterations are made to any of them. Instead, we store we alterations to extra fields set aside for just this purpose. Those extra fields represent the future, and can not have an effect in the present. Therefore, no part of the object that any other thread might be interested in is altered during step 2.

We read one set of data and write to another. We never mix reads and writes, or have several threads write to the same data structure (only one thread writes to a given object's future state information during step 2). Therefore, there's no need for syncronization.

Re:Wow

[GileadGreene]

Now you have to write what effectively amounts to large amounts of multithreaded code - behaving cooperatively on a system with an unknown number of nodes.

Now you either need:
a) A really intelligent compiler
or
b) A really intelligent programmer

or
(c) A language and corresponding underlying concurrency theory that allows you to design and analyze complex interacting multithreaded systems with ease.

Imagine a 16X PCI Express Video Card

[MrRuslan]

I think that would make the GeForce 6800 GT/Ultra/Lite/Whatever look like a Voodoo 1 and a half. a 3d mark pro 05 score of about 30000...hmmm I guess I should patent that...wait a minute...never mind.

Re:Imagine a 16X PCI Express Video Card

[GiveMeLinux]

Imagine a beowulf cluster of those!

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

Read the article you pontificating ignoramus, or are you just trying to seem intelligent without making the effort to actually be so.

Yeah, that's what I thought. You are a worthless dick weed, please go kill yourself and thus vastly improve the mean human intelligence quotent.

Oh yes, please kill your family on your way out as we can do without the gene pool contamination.

In your next life, consider first learning to read, then learning to comprehend what you read, and, lastly learning when & how to appropriately share your perhaps-by-then worthwhile opinion.

Until then, please go back to watching teletubbies whilst sucking your thumb and cuddling your carebear.

Reason why IBM sold PC unit to China?

[kyonos]

Wonder if IBM looks into the future and doesn't see PCs anywhere? Intriguing possibility.

Re:Reason why IBM sold PC unit to China?

[CwazyWabbit]

Or looked at the past and saw their PC unit continually making losses?

Re:Reason why IBM sold PC unit to China?

[Anonymous Coward]

Well, yeah.

Re:Reason why IBM sold PC unit to China?

[Vacindak]

From my understanding, IBM's PC unit was always supposed to be a less seasonal cash-cow that would keep their mainframe business consistently in the black. Basically, mainframes take a long time to build and the number of orders is comparatively quite small. Servers are similarly much lower volume than desktops and laptops.

But diversification only does you some good if the things you do actually earn money, or at the very least, function as a loss-leader. Since the PC unit was doing neither, it made more sense to cut losses and pull out. As far as I know, they're relying on software sales and partnerships to prop up the big stuff in the off-seasons. Seems to be working quite well actually.

multicore, stream-processing, vector-oriented BS

[YE]

While I tend to agree the Cell is an impressive architecture, this article is a steaming pile of B.S.

No cache for CPUs? A breakthrough? Hello! Both PSone and PS2 have the so-called scratchpad, which is what the Cell seems to have: a cache which has to be managed explicitly by the programmer. Breaking news: This is a royal pain in the ass. And calculating bandwidth when reading from this tiny scratchpads makes about as much sense as calculating the speed at which a x86 processor can execute MOV EAX, EBX.

Magically "the OS solves everything", and, in an obvious attempt to automatically get OSS-crowd support (is that "slashdot-trolling" or "slashdot-baiting"?) the triumph of Linux is predicted, because it's portable. Good luck getting the Linux kernel and GCC compiled, let alone running well on a massively parallel array of tiny CPUs without cache.

Re:What's that? Microsoft isn't supporting it?

[arivanov]

You have not read it. It will be on a specific class of tasks. It is similar to modern GPUs. They are faster then 10 opterons on a specific task.

Back to the article. The guy seems to understand hardware, but he does not understand shit about software. Once he got past the first 3 parts he started babbling. Linux on cell, so on, so fourth. If he just read his previous parts he should have hit himself on the head. The only type of linux this can run is mcLinux. There is no memory protection as such. So no Linux, no Windows past 2000, no MacOS past X, so on so fourth.

Similarly, it is all nice and well about cell software beasties making herds by themselves and cooperating on a task. I am going to be a spoilsport and ask a nasty question: Err.. What about a security model? Memory protection? Privilege model for communications? So on so fourth...

To continue on this, the power of a modern general purpose OS is the task switching. How long does it take to load and store the context of the vector processing units? Doing so requires moving their dedicated memory to main memory. This will take ages.

Overall, this is a design similar to Cray 1 initial design. Cray initial design smashed the IBM, DEC (and lesser fish) monopoly on big computing iron to bits. Unfortunately the next thing the people buying the Cray asked for was "can we share this resource between two people?". The answer was provided eventually, but by the time Cray could do all the nifty time sharing and memory management tricks necessary to do this its advantage was no longer phenomenal. And all people who could use Crays for single tasks with manual scheduling actually continued to use it that way. But it did not even dent the general purpose big iron market.

But does it run DooM?

[Neuroelectronic]

Classic doom that is.

iddqd mother fucker.

Cost

[News for nerds]

>Did this guy forget that NVidia is designing the GPU for PS3? If Cell is so almighty,
>why does Sony uses NVidia GPU instead of using more Cells for graphic prosessing?

One possible reason is the cost. When you can save a large area in a silicon die by using a specilized DSP, why do you waste some processing power in a CPU? nVIDIA can provide a reasonablly efficient solutions such as texture units and pipes toward more specific types of processing. Cost is everything, when you manufacture millions of them. At least Microsoft seems to have learned it by the tremendous loss incurred by the Xbox business.

Registers price and future techs for PCs

[Anonymous Coward]

So, it looks really interesting, but the question is really the price, one of the most expensive part in a processor (at least as it was explained to me) is registers. The article says that the cell processors will be able to have huge amounts of registers (128 * 128bits registers per APU times 8 + a couple for the PU). If the price for registers is so cheap, I do not see what will prevent Intel/AMD to add a technology which could use multiple sets of registers like a MegaHyperThreading coming over the 80% of the time waiting for fetches (I believe Sun already used the same trick in is SPARC processors).

The price of memory coming down also means that Intel/AMD will also be able to add a huge on their multi-core processors.

Third thing: concerning power, recent developments and rumors are hinting at partly tickless architecture for next generation processors which would decrease in some significant way power consumption.

Guess the PC is not dead ... yet ;-) it has probably still a couple of tricks to play.

Guillaume

Re:What's that? Microsoft isn't supporting it?

[Neuroelectronic]

yea, you are right. I needs to emulate DOS so I can practice my old-skool Rise of the Triad skillz

Re:What's that? Microsoft isn't supporting it?

[vandy1]

I think that it seems more likely that someone will licence the Cell and replace the G5 core with an x86 core. And I don't see anything wrong with that; it's the normal way to do things :)

So, I think that you'll find that this is about the embarassingly parallel stuff - compilation, for instance, could be implemented on the cell. So what might happen is compilers compile targeting an x86 or G4/5 with APUs attached. All in all, I think that we'll need some pretty sweet compilers for this cool stuff :)

Cheers,

Michael

Obscene

[!the!bad!fish!]

... promises seemingly obscene computing capabilities for what will rapidly become a very low price.

It's for cheaper better porn.

Sounds like a hard sell

[Anonymous Coward]

huahahaha

virus play ground

[Neuroelectronic]

after reading the article, this cell architecture sounds like a virus breading ground. networking without user intervention? across any kind of cell system? imagine a virus spreading from your home computer to your cell phone to your servers at work. just add some random "menes" and polymorphism and you have enough processing power to create a petry dish of evolving computer virii.

Re:multicore, stream-processing, vector-oriented B

[Anonymous Coward]

If it's simply a "programmable cache" like you are so quick to claim then that means you could duplicate traditional cache behavior in software.

And in that case I don't really see where the porting problem comes in.

Re:What's that? Microsoft isn't supporting it?

[halaloszto]

Windows will not run on it, but that does nto matter.

You have to realize, that PCs are no more the high tech in the computer industry. The fastest mass produced CPU on earth was in a PC in the '80s, in the '90s, but not nowadays.

Just like the most PCs were sold for engineering/scientific uses, than this shifted to business administration, and is shifting to households. And will then probably shift to the trashcan. Microsoft was doing all these shifts together with the PC industry, and as the revenue is moving to consoles/dvd players/entertainment centers microsoft is trying to move with it. MS may be successfull. The x86 will die. Applications and sw support was important for engineering, was for office use, is for home/internet use, but is much a simpler question with consoles.

I do not say the PC will die, but it will never more be high-tech. Just like there are very sophisticated alloys, still hacksaws are made of good old steel, because nobody would pay 10 times for the ultra modern alloy. vajk

Re:What's that? Microsoft isn't supporting it?

[BlackHawk-666]

Not just games too, since it is also good for DSP based activities. Audio processing is a DSP heavy task, imagine the quality of the audio you could get out of these buggers. If you can hack it and load Linux on, then you could use it as the most powerful DSP based soft synth around. Think Nord Modular style power and you'll get my drift. I'm drooling just thinking about it - cheap consumer hardware that could provide the horse power of a full rack of specialised audio gear.

25 GFLOPS been already done on multicore......

[Anonymous Coward]

www.clearspeed.co.uk

Unfair comparison

[Stripsurge]

Since the main goal of the chip is to pump through graphics, regardless of what device its in, a GPU is better grounds for comparison.

From TFA: "Existing GPUs can provide massive processing power when programmed properly, the difference is the Cell will be cheaper and several times faster."

Its supposed to do 250GFlops when? 2 years from now? Apparently the Geforce 6800 Ultra will do 40GFlops and thats today.... extrapolate with some doubling here and there it seems a lot more reasonable.

So the big thing is that it comes down to programming. It came up a few times in the article "Doing this will make it faster but will make for one hell of a time for the programmers" It may have a huge potential but may take a while to get everything efficiently as Sony would like. Reminds me of when the GF3 first came out and was beaten by the GF2U in some tests. IIRC it took a while for games to come out that took advantage of its programability. It'll be interesting to see how well the programmers can fair between now and Cell's release.

Re:Unfair comparison

[Hal_Porter]

Its supposed to do 250GFlops when? 2 years from now? Apparently the Geforce 6800 Ultra will do 40GFlops and thats today.... extrapolate with some doubling here and there it seems a lot more reasonable

And a Geforce 6800 probably costs as much as a whole PS3 with the Cell chip.

Mind you, as others have pointed out this article is confusing the peak performance with average performance. And as the chips like the i960 show, that gives a highly misleading impression of what the chip can do.

I'd guess this the PS3 with a Cell and low end Nvidia GPU will end up running the same sort of games as a fast PC with a high end Nvidia GPU but in a much cheaper system. Which is no mean feat when you think about.

Re:Unfair comparison

[Anonymous Coward]

Since the main goal of the chip is to pump through graphics, regardless of what device its in, a GPU is better grounds for comparison.

Agreed. Keep in mind, GPUs aren't really programming by an end user though.

Its supposed to do 250GFlops when? 2 years from now? Apparently the Geforce 6800 Ultra will do 40GFlops and thats today.... extrapolate with some doubling here and there it seems a lot more reasonable.

This article's crap. It's going to do a Teraflop. I don't know the exact release specs but it will certainly do more than 250GF when it's released. This chip is worlds faster than existing GPUs.

It'll be interesting to see how well the programmers can fair between now and Cell's release.

It's actually easier to program than the Emotion Engine so it should make programmers happier than the PS2 did (which of course doesn't say much).

Re:Imagine a Beowulf cluster of...

[Neuroelectronic]

if you bothered to read the article a beowulf cluster would probably just slow down the potential of the cell design. the buss on these is ment to be directly connected to other cells. although it would have problem working over a cluster, it would be limiting, expensive and stupid.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

If you think about it, to the vast majority of users (ie joe public) its going to make no difference.

Does it speed up the internet ? No
Does it make Word printer faster ? No
Does it view p0rn quicker? No

Yes its good but it will be of limited use to Joe Public

Re:What's that? Microsoft isn't supporting it?

[Ash-Fox]

AROS probably could run on it.

Re:What's that? Microsoft isn't supporting it?

[pommiekiwifruit]

That's if they can even build a secure, stable OS with a decent UI)

And boots up instantly and doesn't sound like a Boeing 747 taxiing.

It needs some serious software to work!

[ponos]

There are several assumptions that lead to tremendous theoretical performance figures. The simple fact is that like the Itanium, the Cell processor depends on some rather complicated software that will solve issues like parallelism, coherency etc. The article clearly states that the Cell architecture is a combination of software and hardware (1st page). This is good because performance can always increase (via a better OS or microcode) but it is also bad because it means that initial versions may not stand up to their performance claims.

Also, let's not forget that developers will be unable to keep up, unless some highly sophisticated libraries and languages are made available. I really don't expect the majority of developers to be able to cope with massive parallelism from the beggining (not just 2x SMP or hyperthreading, this needs a totally different mindset).

To sum this up: the hardware will deliver, but the software is a critical unknown in the equation. I have faith in IBM ;-)

P.

Re:It needs some serious software to work!

[Anonymous Coward]

To sum this up:

Go back to playing Quake on your peecee.

"developers will be unable to keep up"
"highly sophisticated libraries and languages"
"cope with massive parallelism"
"totally different mindset)."

Please stop spouting crap you have no clue about.

There just is no end to the number of dumbshits who try to sound like they have a fucking clue with those exact same quotes.

Re:It needs some serious software to work!

[Anonymous Coward]

including yourself, go back to school

Re:It needs some serious software to work!

[argent]

If the first generation simply speeds up OpenGL and DirectX, it'll have a solid market as a GPU.

Re:It needs some serious software to work!

[ponos]

If the first generation simply speeds up OpenGL and DirectX, it'll have a solid market as a GPU.

Only if it can compete with NVidia and ATI offerings at the same approximate price level (the silicon is probably huge so it will cost more than a typical NVidia/ATI product). Do note that Nvidia chips are already "SMP-capable" (in a rough way) by working in SLI mode. Gigabyte has already built a dual GPU card (SLI 2x6600 on a single PCIe card) that appears quite promising. This technology could possibly scale to multi-GPU cards in the future, if competition demands it. I'd probably imagine Cell chips as PCIe "extension cards" for people with special vector-processing needs instead. The GPU market is a tough one...

P.

Re:It needs some serious software to work!

[argent]

the silicon is probably huge so it will cost more than a typical NVidia/ATI product

Why do you think the silicon would be huge? We know how big a 970 core is, and that's almost certainly the biggest part of the unit besides the on-chip [S]RAM, and the GPUs don't get a "pass" there. Besides, if it's cheap enough to use in a console it's not going to be a hell of an expensive core.

Re:What's that? Microsoft isn't supporting it?

[Ash-Fox]

Service contract

Weird... nothing happened.

multiple ps3 designs

[whatsup_will]

from my reading of several websites, i suspect there will be 2+ ps3 consoles. 1 would have all of these cells and do all the backwork, 2 would have 1 cell or 2 and play the game and 3 might just have 1 cell and be used to play ps2 game.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

So, what you're really saying is you have no experience and no future in consumer computing devices.

The future:

iPod or MS Media Center?
Hint: the answer is the first one.

PS2/3 or Xbox?
Hint: the answer is the first one.

Locked Up

[DingerX]

I read all five sections at once, intending to stream each chapter through separate phases from character recognition to criticism. Unfortunately, every time the article used "it's" in a predicative sense, everything ground to a halt.

Fortunately, cell reading meant I hardly noticed the claim that hardware would compete with the x86 because, unlike the x86, cell computers need all their software written for the specific hardware.

I like how "hardware-specific" becomes "OS-independent". Great I can plug my HDTV into my G/Fs "electrically powered adult novelty device", and harness the extra computing power to find out we are really alone in the world. Of course, no firmware will stand in the way.

I'm also surprised that, in pandering to all the OS underdogs in the slashdot crowd (Great day for Apple, since they like G5s; Great day for Linux, since many obsessive-compulsive coders work on Linux projects anyway), he left out a true lightweight OS designed from the ground up for just this sort of multitasking: Amiga OS 4.0. To get something like this to actually work, you'll need more than iPod huggers, OSX preachers or Linux fans. You need genuine madwomen and madmen. You need AmigaOS.

Re:Locked Up

[Anonymous Coward]

To get something like this to actually work, you'll need more than iPod huggers, OSX preachers or Linux fans. You need genuine madwomen and madmen. You need AmigaOS.

So we need a vaporware OS to get the vaporware hardware to work. Got it.

Next up in magic lala-land - learning to fly by lifting yourself in your ears.

He said 250 GFLOPS.........

[Neuroelectronic]

see the article

they have a theoretical computing capability of 250 GFLOPS (Billion Floating Point Operations per Second) [GFLOPS]

load of bollocks

[Anonymous Coward]

It sounds great, can't wait to have a cell system running through my house . . . until he starts using these arguments that don't really lead anywhere?

"The Cell approach does give some of the benefits of abstraction though. Java has achieved cross platform compatibility by abstracting the OS and hardware away, it provides a "virtual machine" which is the same across all platforms, the underlying hardware and OS can change but the virtual machine does not.

Cell provides something similar to Java but in a completely different way. Java provides a software based "virtual machine" which is the same on all platforms, Cell provides a machine as well - but they do it in hardware, the equivalent of Java's virtual machine is the Cells physical hardware. If I was to write Cell code on OS X the exact same Cell code would run on Windows, Linux or Zeta because in all cases it is the hardware Cells which execute it."

Which basically reads "the x86 platform is like a virtual machine, except in hardware, because all the different x86 chips use the same instruction set"

The funniest bit is the 'achieving close to theoretical peak performance'.

I hope to see more bollocks on fox news.

Re:What's that? Microsoft isn't supporting it?

[binary42]

There is memory protection. Read the whole thing. What I think bit you was the fact he said there was no virtual memory... well even then his wording is confusing as virtual memory is just swaping out pages of memory as you need more. This can be done on the Cell. What I think he is talking about it adress translation. Paging hardware must not implement a full LogicalAddr==>LinearAddr==>PhysicalAddr paging/segmentation unit(I have not read the patent myself). He mentions that during runtime the adress must be physical/real and that, when running on an APU, they may be given access restrictions. I must regress though and tell you that I am no expert either. The OS is in for quite a bit of work when dispatching apulets as i can see adjusting addressing and other things will be as interesting (or more) as different scheduling mechanisms are today in current systems. To get a secure system out of this will require protected memory and if i remember correctly the Cell may be capable of running multiple OSs in parallel VMs. This can be explained by considering that IBM has their own software layer that ones OS would talk to (at least the article made it seem that way). Its amlost like having a micro kernel (or exokernel in some ways) that then have real things atached to it. Like linux for example. Linux can already be run in user mode and even ontop of the L4 micro kernel. Linux has shown to be portable enough (along with most good modern software). I would not have any doubt in seeing this happen with IBM.

4 Cell Processors - technical breakdown

[Anonymous Coward]

As the articles mentions, there will be 4 cell processors in each PS3. I have been perusing the patents applied for by Sony and can offer Slashdot readers some technical insights as to what they will be used for:

2 Cells to handle the load of processing Sony's new DRM technology aka "Fuck You - Shut Up & Consume Technology" (TM).

1 Cell to search your home network for MP3s and convert them all to ATRAC, deleting the originals.

The remaining Cell will co-ordinate with other PS3s that it finds on the Internet and launch a huge permanent DDOS attack on OSNews. It is expected that the number of articles appearing on Slashdot will drop to approx 10% of the current level at about the same time.

Okay, I made the last one up, but I can dream can't I?

And as we all know

[Anonymous Coward]

The ever increasing word processing needs are what are driving the development of ever faster PC CPUs and graphic cards....

Oh wait,...

Maybe a processor that can lift Mesa

[Anonymous Coward]

The linux software opengl lib. It is a 100 ton system killer complete copy of a video card done in processor problem when 40 g is tried to be copied by 6g.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

Back to the article. The guy seems to understand hardware, but he does not understand shit about software. Once he got past the first 3 parts he started babbling. Linux on cell, so on, so fourth. If he just read his previous parts he should have hit himself on the head. The only type of linux this can run is mcLinux. There is no memory protection as such. So no Linux, no Windows past 2000, no MacOS past X, so on so fourth.

Err, no. Linux will run on architectures without MMUs no problem. In fact, the PS3's OS will be based on Linux.

Speed issues

[westyx]

I'm trying to work out how he's getting a 4.6 GHZ G5 on a chip along with all the processing units while apple currently can't make their (singular) chips go above 3 GHZ.

Re:Speed issues

[Anonymous Coward]

"a 4.6 GHZ G5"

Huh?

1) Read
2) Understand
3) Comment

Yes, in that order.

Re:Speed issues

[Glock27]

I'm trying to work out how he's getting a 4.6 GHZ G5 on a chip along with all the processing units while apple currently can't make their (singular) chips go above 3 GHZ.

A good 65 nm. process? =)

IBM is making some amazing process breakthroughs lately...

Re:Speed issues

[rpozz]

You would also wonder how the hell they're planning on cooling multiple 4.6GHz CPUs. Surely those things are going to dissipate a serious amount of heat.

Re:Speed issues

[Vacindak]

I have not a whole lot of faith in the 4.6GHz number either. My understanding is that pretty much everyone is hitting a solid barrier at around 4 GHz. (Intel cancelled their 4GHz part recently, for example.) You can build chips that go faster, but the yields drop through the floor and heat goes through the roof. Multi-core seems to be the only obvious answer, which is where all the big players have been headed for at least a year, closer to two now. But multi-core setups need to be more blade-like, less power, less heat, more cores. 4.6GHz, while well within the realm of possibility at 2 years down the road, is probably still going to be excessively warm for multi-core operation. I would have guessed something closer to 3.5, 3.8, something like that. But what do I know? I'm just doing the back-of-the-envelope thing.

Re:Speed issues

[Vacindak]

Or maybe I'm wrong? Hard to say.

Re:Speed issues

[westyx]

sure.
http://www.blachford.info/computer/Cells/Cell1.htm l

has "something like a G5"in the image.

Beneath that, it has specs, including the line "4.6 GHz"

Next time RTFA might be a good idea, huh?

Re:Speed issues

[westyx]

I have no faith in the article at all. It smells like fanboi. I think the cell is going to do well, but nowhere near as good as this guy thinks so.

Best of both worlds?

[Anonymous Coward]

This is a very interesting architecture. It arises quite logically from the 'rediscovery' of vector processing for high throughput, low interdependence instruction streams that GPUs represent. For a long time, home computers didn't do the kinds of things vector processors were good at - they did complex, heterogeneous instruction streams, and the processors evolved to match. You can't run word on a Cray after all. We got microops, RISC, superscalar architectures, multi-level burst filling caches, branch prediction, hyperthreading, all the things that roughly speaking make a stream of instructions that depend on each other's results a lot, don't repeat themselves much, and get information from all over memory go quickly. The thing is, the jobs vector computers are good at started cropping up in home computer loads, mainly for graphics and media type uses. The industry started responding to this with dinkly little SIMD cores in essentially conventional processors, and then we got the GPU, which makes many more compromises to get you greater throughput. Now we have this thing, which is less graphics specific than a GPU but much more vector-heavy than a microprocessor. Most of the 'traditional PC' work, executing code, will be done in the 'supervising' processor I think, which is why it's so fat - you don't need a G5 just to push jobs around after all. Where these vector units come in is for doing 'work units' of the kind of stuff vector processors are good at - 3D graphics, physics, compression and decompression. Basically, all the heavy lifting needed for games and media use. Those who have said it will be very hard to program are correct. If you ran a conventional program on it, you'd get the power of the supervising processor and not much more. You'd have to start looking for these 'work units' in your game, or whatever, and shooting them out to the vector units, along with 128k of all the data they'll ever need, and then pulling the results back when they're done. You'd have to cut the work units small enough so that they were done by the time they were due to hit the screen, or the speakers, which is why there is so little RAM in each vector unit - the jobs aren't expected individually to run for all that long. It is also why the article mentions realtime stuff built in there somewhere, so you know if you're falling behind before the output buffers run dry, and can do something about it. The more of these vector jobs you found, the faster you'd go. You are really looking however at designing a whole program with the philosophy of an OS writer - there will be no abstraction here, and to get the performance you'll have to work for it. The thing is, where has abstraction really got us, we developers? Code is easier to write, and we can write it more quickly and ambitiously than we used to be able to. The problem is that we seem to have 'spent' a whole lot of the new capacity the hardware industry gives us on this one aim, the result being that this 2GHz P4 runing XP in front of me seems about as responsive as a mac classic. Maybe it's time we gave up a bit of our precious abstraction, worked little harder and passed a bit more of that new capacity we get every year on to the users. After all, the first guy to write a game that really does this for this kind of architecture is going to look pretty good, and the guy who comes out the next week with a simple port of the PC version is not. It'll be interesting to see how the game programmers go at getting the best out of this thing. Rumours seem to indicate that they didn't really manage to 'get' the PS2 in this way...

Re:Best of both worlds?

[the_2nd_coming]

you are making it harder than it really is.

Re:multicore, stream-processing, vector-oriented B

[Anonymous Coward]

Linux/OS X won't run directly on the hardware.

IBM has lots of money and time invested in virtualization of hardware. This stuff they been using in Mainframes for 20-30 years and have it good enough that you can run many OSes inside each other without much slowdown or overhead.

Think about what transmeta does with it's running x86 code on non-x86 hardware.

So you have a 'VM'-style parent OS that will take care of all the abstraction that a modern OS needs to be run. Sort of like a microkernel.

But instead of abstracting in hardware like you do with x86 or current PowerPC, you do it with the software 'VM' OS.

So with some modifications, theoreticly current PowerPC gcc compilers should be able to produce code that will run on a Cell. If the cell is setup.

Of course you have lots of optimizations and such to do, but that's what IBM's millions is for.

IBM already does this with it's mainframes. You have bare hardware. You have VM which runs on the hardware, and have proccessing partitions which get a slice of proccessing time and memory. In those proccessing partitions you run your Linux OS. Thats were IBM gets the claim that a single machine can run hundreds of Linux OSes.

The proccessing overhead for something like this would probably run 5-10 percent. You'd have much slower then x86 word proccessing and web browser, but anything to do with 3d acceleration or floating point caclulations ought to scream.

Cell Architecture

[biglig2]

Fairly straightforward I would have thought; use lots of thick stone, big locks, and don't forget the bars on the window.

Re:What's that? Microsoft isn't supporting it?

[djupedal]

Now look what you've gone & done...

Re:What's that? Microsoft isn't supporting it?

[secretsquirel]

hype hype hype, hype hype hype hype. hype hype hype hyiiiiiiiiipe hyiiiiipe hyiiipeeeeeeeeeee...ooooohhhhhh ya

Pipeline?

[Anonymous Coward]

What everyone seems to be missing, is that it is not necessary to paralellize your software, instead you could organize it into a sort of a pipeline, exactly like CPU is doing but, of couse, with much large blocks.
A DIVX codec on a four processing unit, for example, could use one cell to decode quarter of a frame, then handle results to the next cell, and submit next frame to the first one and so on. Decoding would go almost four times faster, wouldn't it?

MmmMmmmm . . . breaded viruses

[Praxxus]

Do you deep-fat fry those, or are they baked?

Re:MmmMmmmm . . . breaded viruses

[Neuroelectronic]

They're half-baked...

No longer true

[Moraelin]

A long time ago, in a galaxy far away, CPU transistor budgets were measured in tens of thousands. You had _barely_ enough of them for either more registers _or_ a more complex decoder, but never both.

This begat RISC. A CISC computer had a more complex in struction set, but that barely left it with enough transistors for a couple of general purpose registers. A RISC computer, on the other hand, went by the mantra "never do in software what the compiler can do for you", so it had an over-simplified instruction set, but then it had enough transistors left for more registers.

In a sense each of the two was too expensive for _someone_. For CISC, registers were too expensive. For RISC the decoder was too expensive. In truth, both were expensive, and the grand unified theory ;) is simply that you just couldn't have _both_.

Fast-forward a bit, and registers are _not_ expensive for CISC any more. You do mention "what will prevent Intel/AMD to add a technology which could use multiple sets of registers", so the answer to that is: they already do. Both have huge register stacks they internally use for renaming. (E.g., when you swap EAX and EBX, the data isn't really copied, but the register from that huge stack which is currently EAX and the one which is currently EBX, get renamed to EBX and EAX respectively.)

Either way, they already have the Cell's 128 registers, and some even have 256 registers. You just don't see them from the outside. (Which is a pity, since compilers could really use them.)

Is there something to stop them from exposing more of those registers to the outside world? Nope. The AMD Athlon 64's (now also addopted by Intel) "64 bit extensions" already do just that: they double the number of general purpose registers visible to the program. That's largely what gives an Athlon 64 the speed boost when running 64 bit code: the extra registers.

Is there something to keep them from doubling or quadrupling them again? From a technical point of view, nothing whatsoever.

What's been keeping them so far is the software backward compatibility. A Pentium 4 still has to run code written for a 486. So whatever changes you do to their instruction set, they must leave the old pre-existing instructions unchanged. And there simply aren't bits left to add the new registers without changing the whole instruction set.

The migration to 64 bits has been such a good excuse to come up with a completely new instruction set, with more general-purpose registers. But such excuses are few and far between.

As for RISC... it died, it lost the battle. Yes, Apple and IBM still use it as a marketting buzzword, but that's it. There are _no_ RISC CPUs still being produced.

The G5 in Macs is simply a CISC with more registers and a better instruction set, but it's CISC nevertheless. It's internal structure is _not_ RISC, and AltiVec is _not_ RISC. They're in fact contrary to everything RISC stood for.

Ditto for Sun's UltraSparc.

(And everyone arguing that a G5 is RISC, has obviously never programmed a RISC CPU before. You had to take care of every single detail in software, because of the mantra "never do in hardware what the compiler can do for you." Even recovering from a pipeline overflow when an interrupt came, you had to do that in software.)

Hope this helps.

Re:No longer true

[Anonymous Coward]

Perhaps you should get someone to explain to you what constitutes a RISC chip before you run off and write your Wiki article.

Fucking 13 year olds...

Re:No longer true

[Moraelin]

Well, then educate me. What does constitute a RISC computer? Take the block diagrams of a G5 and of, say, a Pentium 4 and educated me: exactly what's in a G5 that makes it a RISC, and obviously isn't there in a P4? No, seriously.

The _whole_ idea and advantage (for that time) of RISC was that it basically exposed its microcode to the outside world. (Well, not 100% accurate, but as a metaphor it will have to do.) Any instruction required only minimal decoding to directly drive the ALU and the rest of the CPU.

Exactly how does a G5 still fit that vision?

Except at some time, and after some Apple ads, it became _fashionable_ to use RISC as a buzzword. Everyone had to claim they too were fashionably RISC.

E.g., when the AMD K6 came and had a plain-old micro-coded architecture, they couldn't just call it "micro-coded". They just had to find idiotic marketting euphemisms that boiled down to "yeah, but internally it's RISC-like. Internally we decode those CISC instructions to RISC stuff, you know." (No, it isn't. It's what we've been calling "micro-code" for decades.)

And ever since every single CPU had to do that too. You no longer hear anyone just claiming "micro-code", they now "RISC-ify" the instructions or some other such crap euphemism that a marketroid pulled out of the butt. "RISC-ifying" stuff is obviously more fashionable than saying you use micro-code like in the 60s.

So, yeah, RISC is a paradox. For a dead technology, it sure became a big marketting buzzword.

But anyway, if any of the above is wrong, please do explain how.

Until then, yes, you just live up to your sig: "Fucking 13 year olds". Maybe next time you can read a technical manual first, instead of just throwing a tantrum when reality doesn't match Apple's ads.

Re:No longer true

[Anonymous Coward]

ARM processors were always described as RISC and they recovered the pipeline in hardware on interrupt. It's bullshit to say that even this level of hardware support is "not RISC". RISC stands for Reduced Instruction Set Computer, and has nothing whatsoever to do with complexity or otherwise of the implementation - it has to do with how many basic instructions are supported. What happened later is a bunch of revisionists decided RISC was "dead" and CISC "won" and then redefined the terms to suit their argument.

Re:No longer true

[Anonymous Coward]

There have been quite a few enlightening discussions about the subject on the usenet group comp.arch. Check out google groups. My impression is that your conception of RISC is very limited. Consider this: Pipelining and out-of-order processing are RISC technologies, because they were made possible by the simpler and more straightforward instruction sets. From this point of view, it is easy to understand why modern CISC processors are claimed to be RISC inside.

Re:No longer true

[Anonymous Coward]

Moron. Your car is loaded with RISC processors.
Sorry to tell you, but CISC lost. It is stuck forever on the desktop. Even Intel manufactures RISC (ARM) for the PDA market. ARM, MIPS, PowerPC (yes) are all RISC designs. They are used everywhere off the desktop. Automotive for engine, transmission, chassis control. SOHO routers and WIFI boxen. Printers. Cell Phones. Game machines (except XBOX).
The list goes on and on and on. From a pure count standpoint, RISC outsells CISC in the real world. Open your eyes.....

Re:No longer true

[Moraelin]

Pipelining was possible and used long before RISC CPUs. It was just called "micro-code" back then.

So at some point CPUs just had enough transistors to return to something that had been used on minis and mainframes long before RISC. RISC had _nothing_ to do with it. If anything, RISC was born from micro-coded designs, not the other way around.

In a way, though, unlike other marketting BS, this one isn't entirely inaccurate. That much I must admit. A RISC CPU isn't that different from a micro-code execution unit. RISC was if anything else simply exposing the micro-code to the outside world, and more or less just letting the compiler generate micro-code. So in a way, yes, they _are_ RISC inside. (Then again, so were the mainframes and minis in the 70s, then.)

However, the opposite is also true. The moment you wrap it inside a more complex design with a pipeline, out-of-order scheduler, and and a complex decoder, the whole design (as a whole) is no longer RISC. It may have a RISC core (actually an micro-code execution unit) burried deep inside it, but the whole design isn't RISC any more than a TV is a transistor.

That's what I meant by "RISC is dead". Yes, everyone may have one of them buried deep inside, if you really want to call it that way, but the whole design is wrapped up in other layers that make it really a CISC.

Re:No longer true

[Moraelin]

OK, duly noted. Let me rephras that: "RISC is dead on the _desktop_." Could be that some ancient true RISC design is still produced for washing machines or such :P

Re:No longer true

[Moraelin]

Heh. I must confess that I was actually waiting for someone to try to play the "but it says Instruction Set" joker, and shoot themselves in the foot in the process. Because, see, the instruction set is precisely what makes a G5 for example _not_ be RISC.

Ever programmed a true RISC? See, the matra "never do in hardware what the compiler can do for you", meant that their instruction sets were very basic by definition. The Reduced in RISC stands for an instruction set reduced to the bare minimum that could still work as a Turing Machine.

Among other things, RISC made famous the LOAD-EXECUTE-STORE sequence. Anything conceptual operation that involved more than one of those, _had_ to be broken into the individual steps. (As I've said before, its instruction set was very micro-code like.)

For example, let's take something like "ADD [EBX+1000], EAX" on a CISC. If the x86 were a RISC you'd _literally_ write it like:

MOV ECX, 1000
ADD ECX, ECX
MOV EDX, [ECX]
ADD EDX, EAX
MOV [ECX], EDX

Each and every single building block of the complex instruction, such as loading the data, adding it, and writing it back to RAM, _had_ to be a separate step.

And if your conceptual operation logically involved a loop, e.g., adding two vectors, you _had_ to write the incrementing, comparing and conditional jump explicitly. Because anything else would be implementing them in hardware, instead of letting the compiler do it for you.

What I'm getting at is that _any_ SIMD (single instruction multiple data) instruction set is by definition _not_ RISC.

Now kindly get yourself a copy of the G5's instruction set. And look at it. Especially look at AltiVec. It's SIMD, isn't it? Now how the heck does that count as RISC?

Well, bingo. If it wasn't CISC already, AltiVec did turn Macs into CISC.

So the true revisionists are the clueless marketroids at Apple, who still hang on that buzzword as some mystical quality that their product has. Never mind that the G4 was losing badly in all benchmarks, it's cool because it's RISC.

No, it no longer is. It's a good CISC design. It's good, yes. I wish Intel had an instruction set like that. But it's CISC nevertheless.

Re:No longer true

[Have Blue]

The concept of "recovering from a pipeline overflow when an interrupt came" makes no sense. First of all, I don't know what you mean by "pipeline overflow". If an instruction cannot be scheduled during a particular cycle, it causes a bubble and gets scheduled in a future cycle. None of that operation occurs outside the processor, the only thing the executing program "notices" is reduced throughput. Interrupts are hardly ever thrown from "inside" any processor except for nonrecoverable error conditions like illegal instructions or divide by zero.

Modern general-purpose CPUs are stupendously complex objects that are very difficult to sort into categories that were defined in the 1970s. That doesn't mean the categories don't apply, just that it's not as simple as "It doesn't have memory operands beyond load and store, so it's RISC!" or "It has a vector processing unit, so it's CISC!"

Re:No longer true

[rpozz]

The Sharp Zaurus (and many other PDAs) runs on an Intel XScale, which is without question a RISC processor.

If you read up on the SPARC achitecture, you'll notice that it has a load-store architecture, and is arguably a RISC chip.

Re:No longer true

[davecb]

Moraelin wrote:For CISC, registers were too expensive

CISC also suffered from complex decoders, and required pipelines as long as two or three steps. This was considered scary-complex at the time.

I suspect a modern CISC might push similar complexity barriers...

--dave

Re:No longer true

[SQL Error]

What does constitute a RISC computer? Take the block diagrams of a G5 and of, say, a Pentium 4 and educated me: exactly what's in a G5 that makes it a RISC, and obviously isn't there in a P4?

Addressing modes. And instruction encoding, though that doesn't matter much any more. It's not what's in the G5 that makes it a RISC, it's what's not in the G5. And that's (a) hideously complicated addressing modes and (b) a hideously complicated instruction decoder.

Re:No longer true

[SQL Error]

What I'm getting at is that _any_ SIMD (single instruction multiple data) instruction set is by definition _not_ RISC.
Only if your definition is broken.

Does a multiply instruction make a processor not RISC? How about divide? How about having floating point at all, rather than having to emulate it?

Sorry, but you don't have the faintest notion of the beginning of a clue of what RISC is.

Re:No longer true

[renoX]

I disagree: RISC means *Reduced* Instruction Set Computer, so it means clearly that to it removes some complexity of the instruction set.

And RISC CPU like G5 or Sparc have a reduced instruction set: load/store architecture, less number of addressing mode, etc..

It is true that under the IS, the engine has become more complex (OOO execution, etc.) but the instruction set is still less complex that CPU existing before..

Each new CPU instruction after the RISC invention set is RISC, so saying that RISC is dead is wrong.

But it is true, that *on the desktop* (and nowhere else) compatibility has won over the performance increase gained by the simpler IS of RISC CPUs.

Re:No longer true

[davecb]

Hmmn, that was redundant, wasn't it (;-))

Re:No longer true

[Anonymous Coward]

"There are _no_ RISC CPUs still being produced."

Guess you've never heard of ARM http://www.arm.com/? They've licensed only about a billion of so ARM CPU's floating around in everything from XScale PDA's to Apple's iPod, and countless other devices in between.

Last time I checked, hundreds of ARM licensees are busily churning out millions more on a daily basis.

Re:What's that? Microsoft isn't supporting it?

[Urkki]

• I do not say the PC will die, but it will never more be high-tech. Just like there are very sophisticated alloys, still hacksaws are made of good old steel, because nobody would pay 10 times for the ultra modern alloy.

If only Cell could do what is reguired of a desktop processor... Ie running tens of processes with total hundreds of threads (This WinXP PC currently has 61 Processes and 523 threads running, current memory usage at half a gigabyte) at the same time, with seamless virtual memory with disk swapping and process level memory protection. Cell can't do this, and if it were optimized for stuff like this, it'd have to lose most of the parallel vector processing capabilities, ie the entire point of the processor would vanish.

As it is, Cell has no place in the desktop, except perhaps as a co-processor, provided they convince MS to define DirectX support for utilizing Cell expansion card for vector processing etc.

I'm not holding my breath here...

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

"Cell can't do this"

I assume you are basing your opinion on absolutely no concrete knowledge, because it would be absolutely pathetic if you really have read up on the architecture and spouted such nonsense.

Re:What's that? Microsoft isn't supporting it?

[LarsWestergren]

Maybe you (and others) haven't noticed, but the desktop PC is a deer in the headlights. Game machines will take over before you can say 'service contract'.

Pft. People have been saying this every time a new console generation is coming. When the upcoming Playstation 2 was hyped, some people were claiming it would easily emulate a PC at many times the speed of an x86. When it came, people couldn't take full advantage of the hardware. When they could some years later, PC hardware had surpassed it. Besides, people value the flexibility of a PC. In other words, bs then, bs now.

Nicholas Blachford is an idiot. Please don't read.

[Anonymous Coward]

Nicholas Blachford is an idiot. Do not read any of his articles. Just to give you the best of Nicholas, read his antigravity article and visit his web site:

http://www.blachford.info/quantum/gravity.html

Also, look at the nose pictures of him ;)

http://www.blachford.info/other/me.html

Seriously, the guy has burned most of his sane braincells.

For serious laugh, read his article series 'building the next generation' from osnews. I really got good laughs from that 4 part series.

Also, it didn't take long to spot a totally idiotic statement from todays slashdotted article:

> Parallel programming is usually complex but in this case the OS will look at the
> resources it has and distribute tasks accordingly, this process does not
> involve re-programming.

Here Nicholas misses the core problem of parallel programming. The program algorithms _always_ have to made parallel. The OS can't do it.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

Keep hugging your peecee...

While the rest of the world is moving on to:

iPods, cellphones, Mac minis, TiVos...

Very interesting, but...

[Anonymous Coward]

Does the article's author read tarot cards or tea leaves too?

I'm under the impression that he does.

- Anon.

Re:What's that? Microsoft isn't supporting it?

[Glock27]

Back to the article. The guy seems to understand hardware, but he does not understand shit about software.

This part I agree with. His statements regarding abstraction are just flat out incorrect. Is this going to be programmed in assembly only? I think not...and if not there is significant abstraction involved. The thing that's closest to his point is that multiple *layers* of abstraction tend to add significant overhead. That doesn't mean that program-level abstractions do.

Once he got past the first 3 parts he started babbling. Linux on cell, so on, so fourth. If he just read his previous parts he should have hit himself on the head. The only type of linux this can run is mcLinux. There is no memory protection as such. So no Linux, no Windows past 2000, no MacOS past X, so on so fourth.

There is memory protection if the PU is in fact "something like a G5". IBM would have to be insane not include a MMU, and it has already stated that it's going to build workstations based on the Cell architecture.

All in all, interesting stuff...we'll see how it plays out. :-)

To continue on this, the power of a modern general purpose OS is the task switching. How long does it take to load and store the context of the vector processing units? Doing so requires moving their dedicated memory to main memory. This will take ages.

This, of course, depends on how many cells are in the box (with 8 vector units per cell) and how many tasks need vector units. The main purpose of the vector units in an interactive workstation will be multimedia processing. How many multimedia applications can you view at once? For me, the answer is one. The vector units may be useful for other things like engineering simulation and pattern matching, but once again how many different tasks using those features will be running at once? Plus if the processors are cheap enough to put 4 in a Playstation, one hopes the workstations will have 8 to 32 of them.

Overall, this is a design similar to Cray 1 initial design. Cray initial design smashed the IBM, DEC (and lesser fish) monopoly on big computing iron to bits. Unfortunately the next thing the people buying the Cray asked for was "can we share this resource between two people?". The answer was provided eventually, but by the time Cray could do all the nifty time sharing and memory management tricks necessary to do this its advantage was no longer phenomenal. And all people who could use Crays for single tasks with manual scheduling actually continued to use it that way. But it did not even dent the general purpose big iron market.

Two points. First, this is based on an already successful processor - the Power series. It already multitasks :-) and is used in a wide range of applications. Second, this will be a low-cost part. Crays were a super high-end system, which cost millions of dollars. Your analogy doesn't work.

Re:What's that? Microsoft isn't supporting it?

[LarsWestergren]

Seeing the number of similar angry answers to all posts questioning the hype around the Cell, I'm thinking you are a very persistent but boring little troll behind them all.

Why are you so emotionally attached to a piece of hardware, what is there to be so angry about?

If the cell does all it has been promised, Linux (and probably Windows, though I don't care) will be ported to it. I will notice very little difference to my desktop experience. IBM makes the processor inside rather than AMD or Intel. Fine by me. This isn't a war you know.

End of an era

[Big Nothing]

Does this mean that all our Beowulf-jokes are obsolete?

Re:What's that? Microsoft isn't supporting it?

[Glock27]

Hello, folks. This is the speed in which these guys are talking! This isn't oh, Most powerfull PC in the world bullshit talk like from Apple. This shit is the next level. This is star treck shit, this is your-computer-is-now-HAL, type stuff.

Well, this could very well be the next "most powerful PC in the world" campaign from Apple. :-)

That said, I think one of the killer apps for this could very well be excellent voice recognition. That alone could provide a giant advantage over existing architectures.

Brings back to memory

[Anonymous Coward]

The story on CNN about Sadd*m of Iraq buying the PS2 for his WMD. What will it be for the PS3? Another lie from GW?

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

It always makes me laugh that today's cynics are always the ones who fell for the hype 5 years ago.

Just ignore the hype and look at the patent and make your mind up based on that. History is bunk.

OK, it's theoritically faster than PCs. So?

[master_p]

First of all, the CELL concept is based on massive multithreading. As several other posters have said, software tools suck when it comes to parallel programming. Software engineers suck, too at parallelism.

Secondly, the 4 CELL processors of the PS3 will NOT give it a graphical edge over the PC. PS3 games will not be as impressive as the PC ones. The reason is that graphics will be handled by the NVIDIA GPU, not by the CELLs. But the PS3 will be stuck with a specific NVIDIA model, while PCs will be upgraded to the latest GPUs. Imagine the situation when Half Life 3 will be released: my PC will have a GPU that will be vastly more powerful than the one in PS3, simply because the PS3 one will be a generation behind. It is exactly the same situation now.

So what will this tremendous power be used for? Since the GPU will handle the rendering task, what will the vector units do (the vector units is where the power of the system is)? My 3400+ Athlon already handles HL2 physics at 80 FPS, without even blinking (and with simultaneous downloading and compiling!). Sound can be completely managed by a sound processor.

The only area that I see that CELLs are needed for a game is artificial intelligence; maybe by using neural networks in a game?

Finally, why aren't the GPUs networked like the CELLs? isn't it time to move on a scalable 3d graphics system, where the quality of the graphics is a function of the GPU power? The algorithms for that kind of graphics are already known and tested. adding more GPUs will make the graphics better, revealing details not visible without those more GPUs. That would be a motive for buying a next generation console.

Re:OK, it's theoritically faster than PCs. So?

[nagora]

Firstly: all your points are addressed in the article.

Secondly: anyone that buys a PC to play games on has more money than sense and is quickly parted from the latter.

TWW

Re:OK, it's theoritically faster than PCs. So?

[Astatine]

"Firstly: all your points are addressed in the article."

So the poster sounds reasonable so far, but...

"Secondly: anyone that buys a PC to play games on has more money than sense and is quickly parted from the latter."

Well, that's rich. Some of us don't like trying to "play" games on hardware that's barely up to the lowest common denominator. This is where the consoles always end up a year or so after their launch: with games running at 20fps average and dipping into single figures range, at pitiful resolutions. Have you *tried* playing Knights of the Old Republic on an Xbox?

Ewch. Give me a nice shiny PC for my interactive entertainment any day, for the same reason that I bought an expensive Arcam/B&W separates hi-fi rather than a nasty mini JVC thingy from Dixons... 8-)

Whether or not the PS3 hardware is sufficiently revolutionary to truly offer a PC-beating gaming experience remains to be seen. On paper it sounds up to it, but as another poster has said, so did the PS2 before launch...

Re:OK, it's theoritically faster than PCs. So?

[Siker]

So what will this tremendous power be used for? Since the GPU will handle the rendering task, what will the vector units do (the vector units is where the power of the system is)?

Actually, the CPU speed has a lot to do with graphics speed. If you look at recent performance charts for nVidia's high end GPUs in SLI setups you will find that their performance levels off unless you run the absolutely highest resoultion with top filtering and antialias settings. In fact, the high end cards are still CPU limited at the highest settings for many but the most recent games. [Tom's Hardware Guide]

In addition, programmers will always find things to do with additional CPU power. Ray traced occlusion culling to reduce the number of polygons sent to the GPU is one idea if you have extreme amounts of processing power just sitting around. That in turn would allow you to use extremely advanced pixel shaders as overdraw is almost eliminated. It would also allow you to add a few more polygons to every scene, knowing that most polygons are correctly culled.

Re:OK, it's theoritically faster than PCs. So?

[nagora]

This is where the consoles always end up a year or so after their launch:

As do PCs. It's just a constant upgrade treadmill.

. On paper it sounds up to it, but as another poster has said, so did the PS2 before launch...

And it was, and is. For the price of a PS2 you simply can't get a PC that will play games as well (at all, in fact). The money saved can go on actually buying games for it. The PC has a lot of things going for it, games isn't one of them.

TWW

Re:OK, it's theoritically faster than PCs. So?

[mwvdlee]

Since the Cell processors are basically arrays of vector processors, quite similar to the shader units in GPU's, I suspect NVidia will just implement the specialized low-level 3D stuff and leave all the shader work to be done by the Cell processors.

So basically you'll have a fixed graphics core which isn't likely to change (since it hasn't for the last couple of years) and an extremely flexible and powerful array of shader units.

Re:OK, it's theoritically faster than PCs. So?

[pfft]

So what will this tremendous power be used for?

My guess is that we will see a series of first-person shooters with seriously realistic smoke. Turbulence when you walk though it, gently curling smokestreaks from the muzzle of your gun when you are standing still, vorticity around sharp-cornered objects, convection around hot areas, the works. All computed on the fly in real time, using massively parallell processing. :)

Seriously, while I can certainly see why IBM would want to build supercomputers around this architecture, it will be very interesting to see what the poor game developers will make of it...

Re:OK, it's theoritically faster than PCs. So?

[ultramk]

Secondly: anyone that buys a PC to play games on has more money than sense and is quickly parted from the latter.

They are parted from their "sense?" What?

(I assume tht you're trying to reference the "a fool and his money are soon parted" cliche, but you just don't have the grammatical skills to pull it off.)

Sorry, but I make a decent living at my job, and I bought (built, actually) a PC just to play games on, even though I use Macs the rest of the time. Why? My home machine (Dual G4) still does everything I need to be able to do, and everything I want to be able to do, with the exception of newer games.

Instead of upgrading to a G5 to improve my game situation, I built a PC, which I don't use for anything important. Games only, pretty much.

Just what kind of point are you trying to make? Pick one:
1. Computers are for serious business! How dare you enjoy yourself by using one just for fun!
2. If you have enough money that you can afford to spend more than $5 on something that's just for entertainment, you should donate that money to feed the homeless! You people make me sick!
3. You could have spent that money on something useful, like golf clubs. You fool!
4. How dare you have discretionary income! I work 12 hours a day at BurgerWorld, I should be the one with the sweet, sweet Benjamins!
5. I don't enjoy playing games, therefore neither should you! Ha hah!

Give me a fricking break.

m-

Re:OK, it's theoritically faster than PCs. So?

[ultramk]

The PC has a lot of things going for it, games isn't one of them.

Uh, right. Unless you like RTS, FPS, MMORPG, and the like.

Don't get me wrong, I have an xbox and a PS2 (which I got just to play Katamari Damacy), and I love racing games, but there is more to life.

like WoW, CoH, Half-Life2, UT, etc etc etc....

Not everyone likes platformers.

m-

Re:OK, it's theoritically faster than PCs. So?

[nagora]

but you just don't have the grammatical skills to pull it off.

Well, I typed "latter" instead of "former", if that's what you mean, yes.

Just what kind of point are you trying to make? Pick one:

*List of stupid suggestions snipped*

The point was: for a fraction of the cost you could by a console, indeed all the consoles, and have the money left over for a stack of games. Why would anyone want to spend their money on a machine which will be effectively unsupported in a year's time for five or more times the price of a console with a lifespan of four years?

Give me a fricking break.

If you hadn't spent all that money on a PC you could have bought your own break.

TWW

Re:OK, it's theoritically faster than PCs. So?

[ultramk]

The point was: for a fraction of the cost you could by a console, indeed all the consoles, and have the money left over for a stack of games. Why would anyone want to spend their money on a machine which will be effectively unsupported in a year's time for five or more times the price of a console with a lifespan of four years?

"Effectively unsupported?" In a year? Not sure what kind of crack you smoke, but online games have far, far longer shelf lives than that. The economics of things like WoW require that they run on the same HW for a long, long time. My god, even EQ1 is still pretty damn active.

Of course, as my other post states, if you're not into platformers and prefer MMORPG, RTS, FPS or whatever, with consoles you're kind of SOL.

Besides, I already own the consoles. Most of the games suck.

Hey, how are you liking HL2 and WoW on the PS2? Great huh?

IMO, naturally.
m-

lol, what a joke

[Anonymous Coward]

What are you, some rabid fanboy?

Even a 1ghz processor has MORE than enough power for a media center PC. How much speed do you think it takes to play an MP3 or do DVR functionality? The only think holding back media center PCs is DRM, price, and interface.

Oh, and btw, you may not have noticed, but Sony's game consoles SUCK. By the time they're released Microsoft + Intel + Nvidia are already making games look twice as good. This has always been the case, and probably always will.

The only reason game consoles exist is because they're mass-produced last-generation hardware that can be had for cheaper than a *modern* PC.

Re:lol, what a joke

[Anonymous Coward]

Even a 1ghz processor has MORE than enough power for a media center PC. How much speed do you think it takes to play an MP3 or do DVR functionality? The only think holding back media center PCs is DRM, price, and interface.

That is absolutely spot-on.

Re:lol, what a joke

[Anonymous Coward]

I'm sorry, how many HDTV streams are you encoding into mpeg2 on that 1 GHz while decoding HDTV? I want to see that code.

Re:lol, what a joke

[popo]

Even a 1ghz processor has MORE than enough power for a media center PC. How much speed do you think it takes to play an MP3 or do DVR functionality? The only think holding back media center PCs is DRM, price, and interface.

Wrong.

Wrong.

Wrong.

First off, 1GHZ is more than enough for a linux-based (Tivo or other) media pc. But its nowhere near enough for Windows MCE (which is a dog).

Secondly, even if you wrote lightweight, efficient code 1GHz would stutter like mad on an HD signal.

Thirdly -- no, I have not noticed that Sony game consoles suck. What are *you* some rabid fanboy? There's a year and a half generation gap b/w PS2 and Xbox. And Xbox *sucks* -- it can't do the most basic antialiasing. And Xbox's twin desire to use standard chipsets, and yet at the same time be pirate-proof will be its undoing. No hard drive in Xbox 3 Bwah hah hah hah! Way to shoot yourself in the foot MSFT!

Fourthly -- your point about consoles being cheaper because they're mass produced.... uh... yeah?

Re:lol, what a joke

[Anonymous Coward]

I'm sorry, what kind of normal person gives a shit about that? I want to see your girlfriend.

Re:lol, what a joke

[Anonymous Coward]

Not to mention the fact that the reason Xbox sucks (which can't be more ironic) is that Sony has all the developers on their side and is the industry standard.

Xbox is all about Halo and that's it. The number of good titles for Xbox is hilariously low. Now.. I wonder why it is that all those independent software developers don't like MS?

Re:lol, what a joke

[Noehre]

HDTV is already mpeg2.

Re:lol, what a joke

[Anonymous Coward]

> Xbox is all about Halo and that's it. The number of good titles for Xbox is hilariously low. Now.. I wonder why it is that all those independent software developers don't like MS?

Sounds like they're running a very close race, then. Most console games suck, period.

What's with this pro-Sony crap anyways? Sony is the grand champion of DRM and proprietary products - Microsoft designers read Sony guides on "How To Screw The Customer Over For A Quick Buck" and such. People forgetting the MP3 issue? Betamax?

What is it with this "M$ IZ SO 3B1L!!!11!" garbage that makes people run into the waiting jaws of older, more experienced evil? First IBM, and now Sony! People forgetting the IBM anti-trust issues? Didn't read the article about how DRM will be part of the _hardware_ of these Cells? (I wonder if 'Cell' is referring to the prison that they'll be sending all the pirates and open-source "terrorists" to)

Yes, Microsoft wants to corner everybody with some nasty proprietary solution that'll lock customers in for decades at great expense, but that's no reason to run to a company that will do just the same!

It's time to look for something that's actually _different_ instead of the same tired old crap with a new name.

Re:lol, what a joke

[Anonymous Coward]

People forgetting the MP3 issue? Betamax?

ATRAC/MP3, yes; Betamax, no. Can we please get over the "Betamax failed because Sony didn't license it" crap already? It's a complete myth, just like roughly 90% of the conventional wisdom surrounding Betamax.

Re:lol, what a joke

[lemnik]

Nvidia are the ones creating a graphics processor for PS3. They claim it will be the most advanced, feature rich and powerful graphics processor ever created (to be expected). Need I say more? http://www.psinext.com/itex.php?iid=432

Those who do not learn from history are...

[3vi1]

doomed to repeat it.

I have not RTFA yet, but if programming this thing to make use of the multiple cells is anything like it's been on every other multi-processor system known to man... Sony's going to find developers producing less than the best software for the PS3.

Remember Sega Saturn? The PS1 ate it alive, though on paper the Saturn had more raw processing power. The problem was that developers couldn't find a good way to divide the load between processors. You had one doing practically nothing while the main game ran on the other.

This whole "cell" thing is going to make porting software from the PC a *bitch*, unless the developer is lazy and just uses one cell. Let the crappy ports begin!

-J

Re:Nicholas Blachford is an idiot. Please don't re

[Anonymous Coward]

Yes, because the guy has a goofy website he must be ignored!! He likes linux, too! He must be a complete crackpot and because he is making a joke about his nose all of his ideas must be bullshit!

God, I love being a anal asshole. And my shit don't smell either!!!

This is why there will be no G5 PowerBooks

[rob_osx]

I http://www.tweet2.org/wordpress/index.php?p=13 wrote about this a few days ago. IBM's 970 and derivatives are just too hot for notebooks. Apple doesn't want to spend the money to have IBM develop 2 different processors, one for desktop and one for notebook. Apple wants a chip that has a future, and IBM, Sony and Toshiba have spent billions making sure this chip scales in both power and performance and goes to 65nm. Apple's future is the Cell. Core Image in Tiger is a hint towards the direction Apple is heading. Apple is preparing not to have an Altivec unit, but GPU and APUs.

IBM has said that one rack of Cell servers would have around 16 teraflops of performance. Using Apple's current Xserves it takes 40 racks to get 25 Tflop/s of performance. Apple states that one rack of Xserves currently can produce 630 gigaflops of performance. That means a single Cell server will be able to deliver 25 times the performance of a current Xserve.

No G5 in PowerBooks, in 2006 Apple will announce a whole new lineup of computers, all based on the Cell.

Re:What's that? Microsoft isn't supporting it?

[w.p.richardson]

Windows MCE has already had it rear-end handed to it by MythTV.

My Mythbox is unencumbered by proprietary crap. Can you say the same for your PS3?

Crackpot...

[Dozix007]

Yes, that's right, I said it: Crackpot. The guy who wrote this "explanation" must be having a love affair with the IBM CEO. The computing he is talking about can occur in one of two situations: distributed computing or a "super computer" (which is the same thing). It is IMPOSSIBLE for a single processor to yield that much computing power. The problem is the speed of electricity WILL NOT KEEP UP. And the concept of distributed computing is going to run into the same issues as other networks in the past. This seriously sounds like some cocaine bunny that read papers from about 6 years ago, and possibly the "maybe" speech by Sony.

Re:Crackpot...

[EvilTwinSkippy]

(Disclosure: I'm an Electrical Engineer who actually STUDIED computer architecture and chip design.)

Intel chips are bloated monstrosities. And single chips DO have that much capacity today. The trick is manufacturing them small enough, and efficient enough that they don't require liquid nitrogen to cool them. (Ala the Cray.)

And the speed of electricty is slightly less than the speed of light: 3e8 m/s^2. That's a 3 with 8 zeros, or 186,000 miles per hour. An electron could travel from the sun to the earth in 8 minutes. An electron would circle the globe 8 times in a second.

The distances in a microchip (micro, read the small) are on the order of nanometers. That's 1/1,000,000,000 (1 billionth) of a meter.

You do the math.

Re:Crackpot...

[chadjg]

(Disclosure: I'm a Mass Media weenie, so go easy on me)

Will this asynchronous, clockless computing business solve the clockspeed/propagation issues that are becoming an issue? Will the physical distance between one end of a Cell processor and the other end take the edge off the propagation/timing issues? It looks like a signal would only have to go a relatively small distance before it is somehow packaged/translated and prepped for the interface to the other chips somehow.

It's obvious that I'm underprepared to discuss this, but i'm just trying to tune up my BS filter and it's so damn interesting anyway.

Re:Crackpot...

[EvilTwinSkippy]

Quick lesson on computer architecture.

Clock in a computer act like the drum major in a marching band. Everyone takes their cue from the major's baton on when to start musical notes.

In a clocked computer major steps are broken into what is called the "pipeline". When you add two numbers there are several steps required to load the information into the right places on the chip, adding takes a certain amount of time, and then the result has to be dug out of the adder and fed to the correct output register. Each of these steps take a clock cycle. Since they don't overlap, you can make the processor seem to work 4 times as fast by feeding in new values at the same rate you snatch the results. You just have to know that the answer you are looking for takes X clock cycles, and anything in the register before or after that is not the right answer.

Imagine, if you will, that a chip is a small factory with workstations and coveyer belts. Instructions are passed to the factory by the program through the factory's loading doc (the registers.) The clock is used to coordinate the movement of products between the registers and "workstations" in the chip. Many times the "product" of on operation never leaves the loading doc, it is fed as an "ingredient" for the next instruction/

Clockless computers are designed in such a way that each componant plays whisper down the lane. The results of one step of the operation automatically triggers the next phase of the process. Short operations don't have to wait around for the longer ones to finish before they move on.

However, what you run into with clockless computers are propagation delays. You can never be sure how long an operation is going to take, because each step of the process requires time to perform, and the response time for some operations varies with the input. You also run into problems with traffic control. What if two steps in the process require the same "workstation"? Who goes first? Sometimes you can't tell. So instructions in a clockless chip have to be very carefully designed to prevent contention.

If your system bus isn't clocked you also have to implement traffic signals to control who is getting what for what. Complex, but doable. And all this traffic management after a while starts to act like downtown stop lights. Each intersection is a delay.

So you see, propogation delays have nothing to do with the speed of light, and everything to do with the complexity of the machinery. They are having trouble today with large blocks of RAM, because to address a single cell in a massive array takes time. Picture having to drive, at random, to one of 200 million homes. The roads might take time, but figuring out which way to go at what intersection is what really kills your speed.

Distance sometimes becomes an issue on chips because very thin, narrow, circuit paths are vulnerable to electrical noise. It's not really a speed issue, except that one answer to noise is to slow down. Kinda like speaking over a crappy cell phone. Yelling loader doesn't really help, but speaking clearly does.

4.6 Ghz ? I don't belive it

[S3D]

Article mentioning the core running at 4.6 GHz. I don't belive it. The Intel using every trick in the book to push up frequency, even compromising overall performance, but unable to push above 4 Ghz. Cell seems have as complex architecture as P and multiple cores on the chip and claim 4.7 Ghz ?

Re:4.6 Ghz ? I don't belive it

[rob_osx]

Look at this article and then believe it.

http://www.siliconvalley.com/mld/siliconvalley/103 23259.htm

IBM has made the Cell for servers and embedded applications. I don't know much about the author of the article, but the Cell will change computing.

Here's my analysis on why Apple will use the Cell http://www.siliconvalley.com/mld/siliconvalley/103 23259.htm

Re:4.6 Ghz ? I don't belive it

[rob_osx]

My link to the analysis of Apple's use of the Cell was wrong. http://www.tweet2.org/wordpress/index.php?p=13

No originality

[Anonymous Coward]

I for one quite like the idea of watching "Contact" on my TV while a PS3 sits in the background churning through a SETI@home [SETI] unit every 5 minutes.

Eevry time distributed computing is brought up on /. the only thing anyone can think of is SETI!

My home cluster is setup for 3D rendering and (eventually) real-time video editing with Cinelerra.

oh lord...

[yahkah]

...save us all from this idiot!

seriously, the gentleman that wrote this article is a worthy match for Bowie J. Poag (described properly at http://bowiepoag.is.batshitinsane.com/).

it is completely clear from the text that he knows shit about architecture of computer systems and even much less about parallel processing.

communication vs. processing ratio of parallel algorithms? never mind, we'll use the method at http://www.blachford.info/quantum/fastlight.html for communication between "cells". inherent serial nature of many algorithms? who cares! the OS will implicitly parallelize any algorithm, no matter, how serially designed...

anyway, take a stroll over the rest of his site: his knowledge of physics (demonstrated in the antigravity machine proposal) perfectly matches his knowledge of computers.

weeeeeeeeee

[Anonymous Coward]

taking into account that the OS needs to take care of the in-between-cell networking and that Windsomething has crappy networking support and that IBM recently opened a lot of patents and that sony has previous experience in oficially running linux on their consoles..

DOES THAT MEAN LINUX MIGHT BE INTO SOMETHING HERE? :D ..that was probably just a rethorical question, we all know linux is always into anything!

microkernel?

[constantnormal]

Isn't this kind of architecture extremely well-suited to being managed by a microkernel, with a level of abstraction provided between the details of keeping the processors busy and the dispatching of processes by the OS?

In this manner, the microkernel becomes a sort of "Cell compiler", breaking down the work presented to it and feeding it to however many processors are actually present in the hardware.

The number of "logical processors" seen by the OS might have only a faint resemblence to the actual number of Cell processors under the hood.

Compiler technology - OpenMP

[S3D]

One question which was not addressed fully in the article was how do you compile/test programs for this thing. The answer is OpenMP. OpenMP is mulithreading API wich can hide parallelization from the user almoste completly. It's embarassingly easy to use - only one line of code is enouth to parallelize a loop. All threads creation/synchronisation remain hidden from user. It's extremly efficient too - I was never able to achime the same level of performance if duing multithreading myself.

Re:Nicholas Blachford is an idiot. Please don't re

[Anonymous Coward]

>Here Nicholas misses the core problem of parallel
>programming. The program algorithms _always_ have
>to made parallel. The OS can't do it

I'm not defending his article, but I don't think this "rebuttle" is correct.
An OS could share work between processors automatically, if the environment it provides for software to run in is insulated enough from the hardware.
Think of a JVM. What if you made a JVM that could run across multiple processors? You wouldn't have to reengineer the programs that run in this JVM to take advantage of multiprocessing, the multi-processor JVM would pass on the benefits.

Re:not a new architecture, and it's going to be to

[mwvdlee]

So what? So you'll just create sloppy code and the Cell will still outperform any other consumer CPU by a large factor.

Other than that, I think the chaining of APU's is a pretty new idea. The way the interconnect works is pretty novel too. There's some other nifty bits too; you might want to read the whole article.

Re:PS3 is useless if ...

[Anonymous Coward]

PS3 is useless if there are not optimizing compiler like Visual C++ for Cell and OS like Windows PS3.

open4free © fiction-science or penetration-pacience?

Not deep at all!

[Junks Jerzey]

This is yet another extrapolation from Sony's patent of a few years ago. There's no new information, other than a lot of guessing about competition and marketshare and all that. Tech-wise, all of this is exactly the same information that's been used to write other articles...and there are are the same massive, glaring holes.

Re:not a new architecture, and it's going to be to

[idlake]

So what? So you'll just create sloppy code and the Cell will still outperform any other consumer CPU by a large factor.

Probably not. A lot of code is performance-limited by memory bandwidth and wouldn't run any faster no matter how fast you make the CPU and FPU.

Other than that, I think the chaining of APU's is a pretty new idea.

No; that's one of the standard ways of parallelizing things.

Pretty much this entire design space has been explored before; the "cell architecture" just happens to pick some point in the middle.

Low on substance...

[STratoHAKster]

The author obviously has neither access to any Cell hardware nor any substantial information yet makes the claim that a SETI unit on a single Cell would take 5 minutes to process. Yet in 'references' we read;

5 minutes for a SETI unit? This could be completely wrong... It is based on the difference between a 1.33GHz G4 (6 Hours / unit @ 10 GFlops) and a 250 GFlops Cell, this assumes the SETI client is using Altivec on the G4 at full speed and the PS3 has 4 Cells. I rounded up to 5 minutes to be conservative.

Not a scientific paper. Move along...

Re:Low on substance...

[Politburo]

Not sure what the problem is. Sure, the figure is pretty meaningless, but at least he outlined how he came up with 5 minutes, rather than just throwing it out there as authoritative fact. We should be encouraging such practices, not belittling them.

Re:Low on substance...

[gl4ss]

* Re:Low on substance... (Score:1)
by Politburo (640618) Alter Relationship on Friday January 21, @11:43AM (#11432879)
Not sure what the problem is. Sure, the figure is pretty meaningless, but at least he outlined how he came up with 5 minutes, rather than just throwing it out there as authoritative fact. We should be encouraging such practices, not belittling them.
[ Reply to This ]*

we should NOT be encouraging pulling numbers from our asses. I can promise you one thing: ps3 won't do a seti unit in 5 minutes.

with same kind of idiotical thinking you could take some dsp chip and claim it does seti in 10 minutes...

fanboy article

[egarland]

This was not a technology article. That was a "I for one, welcome our new cell processor overlords.." article.

I don't see anything in the cell arcitecure that would fundamentally make the same number of transistors at the same speed operate faster. I see lots of bottlenecks, IO overhead and wastet transistors. If there is some magical powerful thing that these can do SO much better than the current X86 instruction set and hardware, guess what, it'll adapt.

x86 adapted to RISC being "wildly faster" and, in the end, became better RISC than RISC was by translating more memory efficent X86 instruction onto a RISC backend. It adapted to SIMD (Single Instruciton, Multiple Data) efficiency issues by adding MMX/MMX2/SSD/SSD2 and 3DNow. It adapted to the reality of 64 bit address space and the need for more registers with the new X64 instruction set extensions. AMD and Intel could add cell hardware and instructions too if they offered anything special, which I highly doubt they will.

Verify facts?

[Anonymous Coward]

This article seems blatantly inaccurate. Has anyone seen any of the other things on this guy's web page? UFOs, anti-gravity machines, faster-than-light travel, and so on. He appears to be a crackpot.

Re:Nicholas Blachford is an idiot. Please don't re

[Anonymous Coward]

> An OS could share work between processors automatically, if the environment
> it provides for software to run in is insulated enough from the hardware.
> Think of a JVM. What if you made a JVM that could run across multiple processors?
> You wouldn't have to reengineer the programs that run in this JVM to take
> advantage of multiprocessing, the multi-processor JVM
> would pass on the benefits.

I assume you mean that the programmer doesn't write the program to be multithreaded (hence solving part of the problem), but the JVM makes a multithreaded version automatically.

If you wrote a JVM that moved execution between processors, it wouldn't work well because the JVM would need to do expensive data dependency calculations during program execution, and the synchronization between processors has its cost too. The research community has thought about this for decades without coming up with a great solution.

The only realistic model for the foreseeable future is to program it parallelizable, ie. the programmer solves the problem. This can work very well for very specific problems.

Re:Nicholas Blachford is an idiot. Please don't re

[Anonymous Coward]

The program algorithms _always_ have to made parallel. The OS can't do it.

Never heard of OpenMosix, huh? Fact is, with the multi-tasking nature of most computers today, the OS can do it!

Definitely a fanboy article...

[Svartalf]

However, having 8 independant vector processors in the same package is like having a 9-core box in hand. For general purpose computing, it's not as useful as one would think, but for scientific and engineering computing as well as game uses, it's damned nifty. It'd also probably rock for other multimedia purposes. Imagine being able to crunch video streams with h.264 at a D1 or better resolution realtime as well as decode a similar stream. Instant Video over IP without any special hardware other than the CPU.

Re:Nicholas Blachford is an idiot. Please don't re

[renoX]

You are not totally right: even if your program is compiled in a binary using sequential intructions, the CPU usually can parallelise execution with additional HW (out of order execution, completion,etc..).

But it is true that within a sequential program, the available paralelism is very low, so I'd say that the author of the article is "wildly overoptimistic" over the performance of a cell CPU.

How vector units could help emulating x86 instructions (except for SSE of course) ????

I wouldn't call him an idiot for this article, but a 'believer': his "faith" in the cell is exagerated!!

Specialised computer are great with specialised SW but suck at general purpose computing or at emulating other (different) architecture..

Wouldn't IBM design a JIT for Java for this thing?

[buddhaseviltwin]

Considering how much IBM has invested and banked on Java, wouldn't you think they would try to design a virtual machine that would take advantage of this architecture? I wouldn't expect the JIT to be able to parallelize (???) everything, but I would think it would know how to detect and translate certain segments of code which are easy to translate to a parallel architecture.

I don't know about you, but when I first heard about cell processors (and that fact that IBM was behind it), I immediately began speculating how IBM would exploit this architecture in their server market. This sounds like the sort of thing that will enable them to sell 256 processor monsters running AIX, DB2 and J2EE.

Even if designing to take advantage of this architecture is terribly difficult, just porting your webserver, database server, and transaction will solve the scalability issues for most Web/Client/Server applications.

Apple/IBM vs. Sony/IBM/Toshiba vs. IBM/Microsoft

[Anonymous Coward]

IBM is buddying up to two (or three depending on your outlook) of the most powerfull forces in "entertainment computing" Sony and Microsoft. Apple and IBM (and Freescale) are plugging away at the Power/PowerPC and passing the fruits onto Microsoft in the next gerneration XBOX.

Sony and IBM (and Toshiba) are putting together some of the most esoteric hardware ever to be contained in a consumer computing device and in all cases IBM sits right in the middle of it. Is this cross-fire waiting to kill the giant, or is IBM counting on its size and contracts to protect it?

What happens when Apple says; "Hey! We want to play with CELL in our next generation hardware." Or Microsoft says; "Whats with giving Sony all the good hardware?"

Whats the potential of all these competetors sourcing silicon from IBM and ending up converging on a common, or nearly common architecture? Is this IBM carefully constructing a future monoply? Is it the beginning of the ability to run software from Microsoft, Sony, and Apple on a single (or nearly single) architecture?

Remember the IBM LongTrail CHRP/PREP development boards? A friend of mine who did some Web work for IBM was paid partialy with a dual processor PowerPC 604 development board that contained both an Apple ROM slot (used the beige G3 ROM) and full OpenFirmware (not the cripled version Apple implemented). Only the lack of well develped drivers for the peripheral hardware prevented it from running as a 100% compatible MacOS/WinNT/AIX (it was given to him with all three installed on the hard drive) machine. This sort of attempted hardware convergence is not a new route for IBM.

What do people out there think is going to happen? Are Sony and Microsoft far enough apart in thier long-term goals to keep working with IBM without stomping all over each other in the process? Where does Apple (or even Nintendo for that matter) fit into this? What rumors or conspiracies have people heard?

Very interesting...

[It doesn't come easy]

I believe this development (the CELL processor) is related to IBM's recent sale of their PC business...I will bet that they are counting on the CELL to revolutionize the state of the CPU, and that demand for the old style chips will begin to decline. The timing looks like they struck while the iron was HOT and got the best deal for old tech that they could (plus, it should also mean they can focus more resources on something they think is the future of processors and computing in general).

Wonder if they have also been working to optimize Linux for the CELL processor? I for one will be watching this very closely...

Vector Processing

[Rick.C]

Back in the 1980s I was involved with Amdahl/Fujitsu's vector processor. It was an Amdahl 470 scalar processor (IBM 370 architecture) combined with a Fujitsu vector box. It featured a reconfigurable floating-point register array. The registers could be set up as 1024 32-bit regs, 512 64-bit regs, or 256 128-bit regs. In operation, you set the configuration, loaded the registers, specified the op-code and clocked the FPU. Presto! The operation was performed simultaneously on all the registers. Then you had to store the results back to RAM. Neat stuff!

The weird part was the programming. The scalar and vector units operated independently, much like the Cell does. There were some new synchronizing instructions such as "wait-for-checkpoint" that allowed out-of-sequence operation to proceed up to specified places in the program. Writing code in assembler was a nightmare, but Fujitsu developed a FORTRAN optimizing compiler (most vector work was done in FORTRAN back then) that automatically handled the optimal register configuration and the synchronization. The folks that wrote that compiler have my utmost respect.

BTW, Amdahl sold only a handful of these processors, mostly to European oil exploration companies to process their seismic echo data.

I can't believe how many "it's" there are.

[jrutley]

x86 has won all of it's battles
equals
x86 has won all of it is battles.

Come on people, it's "its," not "it's."

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

> The guy seems to understand hardware, but he does not understand shit about software.

It doesn't look like he really understands hardware either. Too many bold statements about the future- mostly hype and buzzwords. See his article series in osnews:

http://www.osnews.com/story.php?news_id=7676

Has anyone noticed that Nicholas has fixation for mentioning Amiga in any article he writes?

Re:What's that? Microsoft isn't supporting it?

[Vacindak]

And AMD is just starting to make dedicated media chips.

A modern retelling

[Nom du Keyboard]

One Cell to rule them all,
One Cell to find them.
One Cell to bring them all,
and in darkness bind them.

Yes, the Cell processor will rule the world. After all, hasn't IBM been the Dark Overlord for decades now?

Vectorizing compiler?

[Anonymous Coward]

So this article claims that Cell forgets about abstraction and forces the programmer to go all the way down to the hardware. I cannot believe this to be true - even very good programmers have problems writing multi-threaded code, let alone vectorized multi-threaded code that should be suitable for a variable number of cores in assembly (which is actually an abstraction already).

If this is to be the software setup for their otherwise very nice architecture, I think it will fail miserably.

What they need is a vectorizing compiler at the least, for recognizing vector parallelism in algorithms. A vectorizing compiler alone is a very difficult thing to write. Basically it is all about data-dependency - no data-dependency between X variables undergoing operation Y means you can vectorize operation Y to work on these X variables simultaneously. That is theory. Now I have extended the Open64 compiler to do some small vectorizable kernels. Programmers can write a vectorizable loop in many ways and do it in very obscure ways, way beyond the recognition capabilities of a compiler. Writing a GOOD vectorizing compiler is a huge amount of work and even then it will not vectorize everything vectorizable you throw at it.

Next to that you'll want something that maps data-independent (!) tasks on different APU's using these software Cell's.

To me it sounds like a hellish thing to program. Does anyone know what Sony has in mind? I really don't buy the no-abstraction thing, it goes beyond the comprehension of all but the most genius programmers.

A misused word perhaps...wtf

[Nom du Keyboard]

Sony, Toshiba and IBM's new "Cell processor" promises seemingly obscene computing capabilities

I'd hardly call it obscene, though I could certainly apply that word to some of the PS2 games. <G>.

However in a world where Bad now means Good, and many other adjectives are inverted in common useage, who know wtf the author actually means anymore.

For that matter, who knows what wtf means anymore.

PARAGRAPHS!!!

[Smilin]

Isupposeyoucouldhavewrittenitwithoutspacesandthatw ouldhavemadeithardertoreadbutImnotreallysuresinces uchalargereadwithnoparagraphswassounsightlythatIch osenottoevenlookIbetyouforgettochangehtmlformatted toplainoldtextandfurthermoredidn'tusethepreviewbut tonDon'tbeupsetifsomeoneflamesyou.

6 OF THIS OR HALF DOZEN OF THAT - by distantbody

[Anonymous Coward]

half way through the article i was like 'wow, i cant F*@kin' wait to get my hands on a ps3'...and then it struck me. PR, its just an STI-loving PR'ist. its amazing how subliminal and disturbingly effective some multinationals' advertising can be.(yes, i know
STI aint a multinash, its group)

and on a different note, im somewhat of a ps2 fan,
but i would NEVER let myself be fooled into believing that the warm and fuzzy feeling i once used to get as an materialistic teenager from hearing or seening "SONY" or "NIKE" somewhere was anything more than millions of dollars of PR weaving its evil work...

in addition, if the articles claims are true and STI did end up as the futures equivilent of wintel, dont think that it would be computer-geek nervana, because as i often remind people, the only goal of a corporation is to make money, and wether it be microsoft, sony or anyone else dominating the computer business, you will always be seeing ridiculous prices, por quality, and worst of all, YOU WILL ALWAYS BE SUBJECT TO ATTEMPTS TO TURN YOUR RIGHTS INTO A COMMODITY FOR THE SAKE OF PROFIT, ala microsoft

-distantbody

The real question for Apple...

[argent]

Will the cell-based GPU in the Powermac G6 be labelled "ATI" or "nVidia", or will the Wheel of Life return the GPU to the core again, and it'll be running on the "1.5-way" (G6 + Cell) Power PC daughtercard?

x86, Apple etc Vs Cell my arse

[theolein]

I'm not actually surprised that so-called journalists, especially the technical kind, get good salaries. If you look at the painful clowns running the show at ZDNet, and most technical publications for that matter, including such wonder rags, such as the Register, you know that the Agenda is almost the most important thing. The actual realities of the tech world be damned as long as you have someone passing you your monthly wad of cash.

And this story is no different.

As many have noted, Sony did exactly this kind of hyping the last time around when the PS2, with its emotion engine, was supposed to be the future of all things computing. As everyone knows, the PS2 was a real pain to code for, and the actual performance was not better than the PC's of the day. The Cell will undoubtedly suffer from the same problems when it comes to coding real applications. Concurrency and parrallelism do not an easier coding experience make.

I have no doubt that this thing will be good, but I absolutely doubt that it will have much or any effect on the x86 world of computing. The G4 processor, when it came out with the Altivec SIMD processsor, which was apparently better than SSE at the time didn't turn Apple into the next Microsoft overnight either, did it?

So, I expect that the x86 world will continue to thrive and that Apple will stick some of these Cell processors, having as they do a PPC 970, aka G5, in their core, in some of their machines and will make the usual wild RDF claims about how hot it is while it will be used by only a small fraction of actual Mac developers in reality, the Mac having to maintain backward compatibility only slightly less then the x86 world does.

In other words, it'll be business as usual.

Re:x86, Apple etc Vs Cell my arse

[argent]

I don't know anything about your arse, but this would make a hell of a GPU... and who cares how hard that is to program, you hide all that under OpenGL or DirectX...

Re:x86, Apple etc Vs Cell my arse

[DaveCBio]

Why would you use a general purpose CPU like this for a GPU? You'd get far more mileage out of dedicated hardware and even if the API hid everything it would chew cycles to translate what would be native on a GPU.

Re:x86, Apple etc Vs Cell my arse

[argent]

Why would you use a general purpose CPU like this for a GPU?

Um, because it's not a general purpose CPU? That's the whole reason people are complaining about what it's doing to their arses.

Re:Nicholas Blachford is an idiot. Please don't re

[fitten]

Lots of people have been working on auto-parallelizing compilers. The idea is to take existing code that isn't parallel and during compile time (or run time) make those decisions intelligently and speed up processing. So far, there have been zero successes at it without explicit user directives to tell the compilers where good targets for parallelization are and how to do it (specifically creating threads and/or marking loops that can be parallelized).

If you (or anyone) can solve this problem well, you'd be famous and wealthy beyond the dreams of avarice (assuming you patent it and license it out :))

Re:What is the OS of PS3?

[Anonymous Coward]

Koyto OS for PS3? hehehehe.

open4free ©

Re:What's that? Microsoft isn't supporting it?

[anothy]

the most "interesting" operating system i know of, Inferno, doesn't make use of a MMU. it's designed to be able to run on very minimal or embedded-style hardware (or as an application on top of another OS). it's got some really fascinating characteristics, and its handling of networking in particular is still way ahead of anything else out there, even eight years after its initial beta release. your comments about performance are still entirely appropriate, but there's plenty of very interesting - and even general purpose - things to be done with this.

There's more parallelism in heaven and earth...

[argent]

There's more paarallelism in heaven and earth than are dreamed of in your philosophy...

inherent serial nature of many algorithms

The Power PC 970 isn't a slouch when it comes to serial algorithms. But in a modern computer system, like Mac OS X, there's an awful lot of stuff that's deeply vectorisable. ALL the graphics, for example.

Powermac G5: dual G5 processors with Altivec + high-end ATI GPU, running a heavily 3d-accelerated GUI based on OpenGL.

What would a cell do for this?

Powermac G6: single G5/G6 processor with a cell coprocessor driving a dumb frame buffer, with OpenGL implemented as a stream of cells, Quartz Even-More-Extreme and real-time raytracing...

PS3 with Cool'n'Quiet Technology for Power Saving?

[Anonymous Coward]

10:00 - 11:00 - 12:00 - 13:00 - 14:00 - 15:00 ...
140 W - 10 W - 70 W - 20 W - 140 W - 140 W ...

open4free © (;

What gets me...

[xeon4life]

What gets me is how my computer is consistently outperformed by these gaming consoles. How can a $150 gaming console with 64MB of memory and some 1Ghz Celeron outperform my $2000 powerhouse in graphics processing? There are only a few possible explainations for this paradox:
1) The consoles are very well engineered. Or as made apparent by my computer science teacher, "Just the Apple computers, they have control of the hardware, so they only have to focus on what hardware they allow."
2) The drivers are very well engineered and specifically optimized. In the words of my computer science teacher, "They can optimize the crap out of them."
3) The underlying operating system is far less bloated.

I'm all for extendibility and customization when it comes to our workstations, but if our operating systems, drivers, and 4Ghz Athlon FX, ATI Radeon X850 PCI Express powerhouses are being outperformed by gaming consoles, something has GOT to be wrong...

Re:What gets me...

[HellKnite]

What's wrong is called "Windows" - you hit the nail right on the head with point 3. 1 and 2 are probably true, but don't have as much impact on the statement you're making. A console doesn't HAVE an underlying OS, there is only the game software executing instructions built into an API which talks directly to the hardware. If you look at the XBOX and it's myriad hardware and software hacks, you'll see that you can never run one program with another one at the same time. Certain BIOSes are attempting TSR (Terminate and Stay Resident... remember that from the DOS days?) applications, but when a game thinks it has 64MB of ram to use and it starts stepping on this TSR app running in the background, you get a big mess.

Windows (or any OS for that matter) on the other hand is doing a billion things for your computer while you're playing a game. IT is controlling the TCP/IP stack, IT is dealing with all of your applications you left running in the background, IT is dealing with tons of random drivers which may or may not be optimized for a gaming experience.

Theoretically, if you could make a game on a boot CD which included drivers for all of your hardware, and ALL it did was run that game, you'd squeeze a lot more out of your system than running through windows. However, this is exactly what the article discusses in reference to abstraction... to do the "Game on bootable CD" concept, the game developer then has to code all of their own drivers to accomodate all of the random PC hardware you may have, including video cards, network cards, joystick/gamepads, on and on. So, for the price of slowing things down a little and access to DirectX or SDL, we lose some of the potential of our powerhouse computers by running an OS before we run the game.

Re:What gets me...

[gl4ss]

well. OR THERE IS number 4 explanation: the consoles actually have crappier graphics and are NOT outperforming the pc's.

which would actually be the true answer as well. put them side to side and you'll see it.

Re:Best of both worlds?, now formatted!

[Anonymous Coward]

It still is very bathery, but at least your eyes won't get lost in sameness of the text.

This is a very interesting architecture. It arises quite logically from the 'rediscovery' of vector processing for high throughput, low interdependence instruction streams that GPUs represent. For a long time, home computers didn't do the kinds of things vector processors were good at - they did complex, heterogeneous instruction streams, and the processors evolved to match. You can't run word on a Cray after all.
We got microops, RISC, superscalar architectures, multi-level burst filling caches, branch prediction, hyperthreading, all the things that roughly speaking make a stream of instructions that depend on each other's results a lot, don't repeat themselves much, and get information from all over memory go quickly. The thing is, the jobs vector computers are good at started cropping up in home computer loads, mainly for graphics and media type uses.

The industry started responding to this with dinkly little SIMD cores in essentially conventional processors, and then we got the GPU, which makes many more compromises to get you greater throughput. Now we have this thing, which is less graphics specific than a GPU but much more vector-heavy than a microprocessor. Most of the 'traditional PC' work, executing code, will be done in the 'supervising' processor I think, which is why it's so fat - you don't need a G5 just to push jobs around after all. Where these vector units come in is for doing 'work units' of the kind of stuff vector processors are good at - 3D graphics, physics, compression and decompression. Basically, all the heavy lifting needed for games and media use.

Those who have said it will be very hard to program are correct. If you ran a conventional program on it, you'd get the power of the supervising processor and not much more. You'd have to start looking for these 'work units' in your game, or whatever, and shooting them out to the vector units, along with 128k of all the data they'll ever need, and then pulling the results back when they're done. You'd have to cut the work units small enough so that they were done by the time they were due to hit the screen, or the speakers, which is why there is so little RAM in each vector unit - the jobs aren't expected individually to run for all that long. It is also why the article mentions realtime stuff built in there somewhere, so you know if you're falling behind before the output buffers run dry, and can do something about it.

The more of these vector jobs you found, the faster you'd go. You are really looking however at designing a whole program with the philosophy of an OS writer - there will be no abstraction here, and to get the performance you'll have to work for it. The thing is, where has abstraction really got us, we developers? Code is easier to write, and we can write it more quickly and ambitiously than we used to be able to. The problem is that we seem to have 'spent' a whole lot of the new capacity the hardware industry gives us on this one aim, the result being that this 2GHz P4 runing XP in front of me seems about as responsive as a mac classic. Maybe it's time we gave up a bit of our precious abstraction, worked little harder and passed a bit more of that new capacity we get every year on to the users. After all, the first guy to write a game that really does this for this kind of architecture is going to look pretty good, and the guy who comes out the next week with a simple port of the PC version is not.

It'll be interesting to see how the game programmers go at getting the best out of this thing. Rumours seem to indicate that they didn't really manage to 'get' the PS2 in this way...

IBM decided to get as crude as possible.

[argent]

On the contrary, it's a simpler architecture than the Pentium.

If they think you're crude, go technical.
If they think you're technical, go crude.
IBM is a very technical company,
so they decided to get as crude as possible.

There's no cache, so no cache coherency issues, and coarse-grained memory addressing. The most complex component is the Power-PC 970 core, and it's only a controller... it could easily be half-clocked at 2.3 GHz without significantly hurting the throughput of the vector units... and IBM already has 2.3 GHz 970s in production.

Re:How about quotes from the story submission?

[Psykechan]

[Cell technology] is a revolutionary step forwards in technology and until now, the most serious threat to x86.

OK, untill now it was the most serious threat to the x86. What is the most serious threat now that we have more information on Cell?

Say this 3X fast

[Nom du Keyboard]

If all goes well will Dell sell Cells?

Try saying that fast three times.

Re:Say this 3X fast

[EvilTwinSkippy]

A Dell with a Cell is a swell sell.

This is baseless hype and marketing, nothing else.

[i41Overlord]

I remember before the PS2 came out, they said it was going to have the power of a "supercomputer" and it would render lifelike images in real time. Looking at its graphics would be like watching a movie they said.

They even showed screenshots and videos which showed highly detailed images that looked fantastic. Everyone wanted to buy one.

Yet when the PS2 came out, those fantastic videos turned out to be nothing more than pre-rendered scenes that the PS2 was playing like a DVD player. It wasn't rendering those scenes in realtime. We were all tricked by clever marketers. The actual gameplay that the system was processing in realtime was ok, but nothing out of the ordinary. The PS2 turned out to be just another video game system surrounded by a lot of senseless hype.

I'm not holding my breath for the PS3 and it's Cell processor. It will not have the power of xxx high end computer processors, and it won't be revolutionary. It will just be another evolutionary step in videogames with improvements- but not groundbreaking improvements- in gameplay and graphics.

Let's quit the senseless hype and keep things in perspective.

He must have a Jump to Conclusions Mat!

[rbegga]

For all those who keep saying RTFA to the posters who see this article for what it is: I have RTFA and there is certainly NOT enough information to draw ANY type of conclusions with regard to performance of this architecture, other than that the author's "interpretation" of the patent looks good on paper. Present the article to any other computer and systems engineer and they will tell you the same. I deeply dislike the fact that the Internet has become a breeding ground for speculation and FUD such as this. Can we please investigate sources before posting this stuff? This man has produced an articles about 1.) how to develop a gravity drive while taking into account all the problems presented in Star Trek 2.) Claims to have solved several problems relating to time travel, while admitting it's not possible 3.) freely admits he is not qualified to speak about anything at all and uses Issac Newton as his excuse for taking editorial license. I think everyone is entitled to an opinion, but that it shouldn't be published in a location where impressionable readers can be mislead, or propagated by people who don't know any better.

Imagine SETI in Real-Time

[Nom du Keyboard]

a PS3 sits in the background churning through a SETI@home [SETI] unit every 5 minutes.

Imagine that. A network of PS3s processing SETI data in real-time. Wouldn't be surprised if that application isn't already under development.

And so what if the author is off be even an order of magnitude in his estimate of processing time as another poster suggests due to the lack of actual hardware to test this on yet. Within a couple years of launch there are going to be tens of millions of PS3s out there, all with network connectivity.

Re:Imagine SETI in Real-Time

[DaveCBio]

You are missing on very important point. How many gamers are going to allow their spare cycles to be used for something like that? The vast majority will be offline and those that are online will be turned off after the gaming is over like any console.

Re:Imagine SETI in Real-Time

[Nom du Keyboard]

How many gamers are going to allow their spare cycles to be used

Even gamers sleep ... eventually.

Re:Nicholas Blachford is an idiot

[diamondsw]

Not going to defend Nicholas, but it is certainly the case that better APIs that are highly parallel can be provided by the OS for graphics and AI operations. After all, that's where the bulk of processing time goes in modern games. A lot of work will still be required on the developer's part, but I can see a good chunk bein abstraced behind an API layer.

Re:Apple/IBM vs. Sony/IBM/Toshiba vs. IBM/Microsof

[argent]

What rumors or conspiracies have people heard?

Only Bigfoot's orange juice and eggshell wonder diet.

Re:What's that? Microsoft isn't supporting it?

[Anonymous Coward]

So no Linux, no Windows past 2000, no MacOS past X, so on so fourth.

Uh, Windows before 2000 had memory protection... IIRC it just didn't do the copy-on-write that NT does for dlls, and it allowed you to unprotect shared kernel space and then modify it.

Win9x, which wasn't based on the NT kernel, would be effectively impossible to port to another architecture anyway. It wasn't designed with that in mind and it's so old now that it's not worthwhile.

Easier problem than Itanium

[argent]

The algorithms needed to make the Itanium go fast are still being figured out.

The algorithms needed to vectorise operations are well known and in wide use in every GPU in the world.

You're falling for the hype- hook, line, sinker.

[i41Overlord]

Sony made all kinds of outrageous claims before the PS2 came out.

http://csmonitor.com/cgi-bin/durableRedirect.pl? /d urable/1999/03/25/f-p13s2.shtml

They said it was going to be able to render "Toy Story" like animation in real time.

http://www.time.com/time/asia/magazine/2000/0320 /j apan.sony.html

"But it's souped up with extras: a microprocessor as powerful as a supercomputer and ports for hooking up cable TV, keyboard, mouse, digital-video camera and modem card. The possibilities are huge. "

From just looking at a simple top-level diagram of the Cell architecture, it is clearly shown that the Cell is much more powerful than any other processors currently available.

LOL. You can tell what the real-world performance of an unreleased chip is going to be, just by looking at a simple top-level diagram?

I can guarantee that the Cell is much more powerful than any AMD and Intel processors.

Hahaha. You can guarantee, huh? So the chip in a game system marketed towards gamers is going to deliver more power than high end CPU's that cost many times what the entire game system will cost?

I don't know what you're talking from, but you're not talking out of your mouth.

People tried something like that:

[i41Overlord]

http://www.geek.com/news/geeknews/2003May/gee20030 528020156.htm

The AMD64 port is...

[WillerZ]

Called "Windows XP Professional x64 Edition" and the preview release works pretty well. I'm able to play Medieval:Total War on it quite happily.

Phil

Easier problem than Itanium-Vectors.

[Anonymous Coward]

"The algorithms needed to vectorise operations are well known and in wide use in every GPU in the world."

Not quite. It's still hard to forsee what can be vectorized, and what can't. Especially in a dynamic language.

IMHO The vectorizing decisions should be pushed into hardware. Same with threading.

Re:Easier problem than Itanium-Vectors.

[argent]

It's still hard to forsee what can be vectorized, and what can't. Especially in a dynamic language.

There's some things that are easy to vectorise, and modern computers spend a lot of time doing them. Implement those algorithms for the cell, and package them up to call from applications.

Re:Nicholas Blachford is an idiot. Please don't re

[woah]

How vector units could help emulating x86 instructions (except for SSE of course) ????

I'm not defending or bashing anyone here. But I think the way this is possible is thourgh JIT. Compilation is a parallelisable problem. As you execute code on one cpu you can compile code on another.

The main difficulty would in be the fact that the APU and the x86 architectures are completly different. This might impact performance. Or perhaps it's possible to use the APUs for compilation and the main CPU for execution, or maybe a mixture both approaches.

Anyway, it's way to early make any conclusions.

Compiler technology-Ars-Technica

[Anonymous Coward]

Ars Technica covers the 'why' of the design.

Re:Compiler technology-Ars-Technica

[faragon]

Yes, I saw the same article few years ago. I got the PS2 Linux devkit for a while, and in fact, it is hard to deal with the VUs (vector units, see the Ars Technica refered in the parent).

Here is a resume of why of that architecture:
- With a ~300MHz core the target is to achieve 6 GFLOPS (about the same as a P4 @3GHz).
- You can not afford easily to give 6 GFLOPS with a single CPU running at ~300MHz.
- Then, you have to choose a design with available technology.
- Use a quite common MIPS CPU (R5900 MIPS III, and near MIPS IV compliance).
- Add two vector units with massive FMAC throughput.
- Choose a good bus design for I/O.
- Choose a GPU.
- That is, you have a PS2, state of the art computer architecture.

Of course it is not as easy as to deal with a single and powerful CPU with a "intelligent" GPU attached, but it is quite optimal for a 300MHz core... could you have done it better? ;-)

How could they possibly do this cheap?-Spread 'em.

[Anonymous Coward]

"I'm willing to believe that a 4.6 GHz chip with 8 ALUs and high bandwidth memory would be fast, but even in bulk, there's no way they can afford to put 4 of them in a sub-$500 game console."

Since you've been reading the PR. You already know they intend for the Cell processor to be used in more than just a $500 game console.

Reason why IBM sold PC unit to China?-OS/2

[Anonymous Coward]

That and it was a good way to get in a good poke at MS. They still haven't forgiven them for OS/2.

Plus now they have a gateway into China.

What he really means

[Anonymous Coward]

The Cell has 4 FPUs times 8 PEs at 4.6GHz, for a theoretical 147 GFLOPS with unknown power consumption. The ClearSpeed chip has 2 FPUs times 96 PEs at 250MHz, for a theoretical 48 GFLOPS with 5W power consumption.

Combine 3 or 4 ClearSpeed chips and you'll get the performance of a Cell unit with no doubt lower power consumption. Of course by the time the Cell sees the light of day in 18 months, it is quite likely that ClearSpeed will have updated their chips to match.

aQazaQa

Re:Nicholas Blachford is an idiot. Please don't re

[Anonymous Coward]

> Never heard of OpenMosix, huh?

To take advantage of mosix, one has to program task into _separate_ processes. Hence the programmer can only parallelize on process level (very coarse granularity), and by doing that the _programmer_ solves the problem. So this isn't a good counter example.

Cell Processor article inaccuracies

[Anonymous Coward]

The article is full of inaccuracies (at least 50% of his "facts" are wrong). Ironically his conclusion section is probably correct. (Have to post anon due to NDAs :-)

PC vs Cell ... cell will be assimilated

[dtjohnson]

The article pitches this as a cell vs PC fight where the cell will be 10x faster than the PC we use today and therefore we will all purchase cells. Uh-huh.

When will these kind of guys get it? It has taken years of time and zillions of hours to get to where we are today. The new 10x faster whatever pales into insignificance if it can't do what we are doing right now. The cell approach obviously has a lot of potential but it will never get mainstream adoption unless it integrates into what we already have. If the promise of the technology is actually realized (and that's a big if), it will reach the PC platform only as some sort of transparent add-on to the current platform architechture. Then it will be gradually integrated until new PCs have a sticker on the front that says 'includes CELL' or whatever.

We are software limited, not hardware limited, and anything new had better work great with everything we are doing now. Otherwise, that new standalone CELL thingy will be just another piece of really fast hardware that everyone ignores because it can't do anything.

COSA.

[Anonymous Coward]

Well I'm going to get this guy Slashdotted.

The COSA OS. Try matching that with the CELL?

Re:COSA.

[Anonymous Coward]

Papers trying to apply biological metaphors to computer systems usually bring the phrase "a little knowledge is a dangerous thing" to mind.

Thanks a lot

[Raunch]

Cell provides a machine - but they do it in hardware, the equivalent of Java's virtual machine is the Cells physical hardware. If I was to write Cell code on OS X the exact same Cell code would run on Windows, Linux or Zeta because in all cases it is the hardware Cells which execute it.

So the apps will run on anything- you merely need to rewite every OS ever made!

Re:Nicholas Blachford is an idiot. Please don't re

[fizban]

But once you've compiled it, you don't need to compile it again, therefore your JIT optimization is over within a fraction of a second.

..it's going to play German heavy metal.

[javaxman]

To quote TFA :This system isn't just going to rock... it's going to play German heavy metal.

Or, rather, from the sound of it, programming it is going to be like listening to German heavy metal. As in, really, really hard...

Uh, yea, it's important to keep in mind that this is all pretty well theory, I could find some other interesting patents that describe in detail that flying car we're all supposed to have, along with thousands of other undeliverable advances. On the other hand, there will be a PS3, and Sony, Toshiba and IBM are spending billions to make sure it's a super-fast successful system, so... we'll see.

was the ps2 the supercomputer it was said to be...?

Actually, uh, it's how old and does what for what price and has been how successful ? I'm not sure I'd scoff at the PS2, no matter how much marketing drones over-hype products.

OMG

[Anonymous Coward]

after seeying the arquitecture and the teoric numbers, the Cell is like an arquitecture that I was thinking about in a OS.

PU -> DMAC -> VARIOS APUs
EXOKernel -> Security Policy -> Hurd-like servers

massive parallel

[Anonymous Coward]

how does the function
software_difficulty( number of CPUs )
look like?
Is 100CPUs more difficult then 10CPU
10CPU is certainly more difficult than 1 CPU

Sounds like hot air, but maybe not

[Rize]

You would think that we would have learned our lesson from the PS2 hype, but then again, perhaps Sony has learned there lesson as well. I think one thing is clear. The PS3 will easily outperform the XBox 2 and Nintendo's "revolution". As long as the programmers can cope with the new paradigm.

Re:What's that? Microsoft isn't supporting it?

[popo]

I hear you. (and totally agree). I'm comparing encumberance b/w Windows MCE and the PS3. Ultimately both will look encumbered as hell next to opensource linux-based DVR software.

(Building my Myth TV box now...)

Unanswered question

[Locke2005]

What is the power consumption? Will Sony be able to build the PS3 without a fan, or will customers be able to use the PS3 to heat their whole house in the dead of winter?

Re:How about quotes from the story submission?

[Anonymous Coward]

AMD64

re: x86

[web_boyo_in_sac]

given the modular design of Cell, couldn't an x86->Cell emulator module be built?
Hardware emulation, on a hardware level.
The module would talk the x86 arch to the outside, but talk whever Cell uses on the inside.

Would allow me to run every x86 app I have from winders to suse, but I wouldn't have to port anything to a new arch.

might be a little slower, but adding more modules could make the cell cluster look like multiple CPU's to an OS.

unless I'm missing something fundemental about Cell's design.

Re:What's that? Microsoft isn't supporting it?

[Deliveranc3]

Two words

PSX Three


These companies have been screwing the pooch on this one for a long time and they aren't going to stop anytime soon.

Bittorrent is it, bittorrent is video on demand.

That cable in your wall will be digital tcp/ip.

Re:What's that? Microsoft isn't supporting it?

[Locke2005]

I have a Sony Media Center PC (Running Windows XP Media Center) and it is a complete piece of shit. (Since it is always recording and playing back video at the same time, it appears to drops frames like crazy. Also, Doom 3 will only run with a decent frame rate at 640x480. All this for only $2400!). I'm unclear whether this is Sony's or Microsoft's fault (probably a combination of both), but it is abundantly clear that Windows Media Center is not yet ready for prime time. Hmm... does the Linux PVR software support the latest Sony PC hardware?

Dreamcast

[eobanb]

Actually, a lot of what killed the dreamcast was that it is trivial to pirate games for it.

Wow

[TakeMyName]

Wow, this article is amazing, Joseph Goebbels would be proud. Why does this rediculous sort of hype happen just with playstation?

Think another way

[marcus]

The current processing bottleneck, and the reason for caches in the first place, is the bandwidth between the processor and the memory. A "normal" memory bus cannot keep up. This is why you see so many attempts to speed this particular part of the system up. There is RAMBUS, DDR, even HyperLink.

What these guys are trying to do is move the processor to the memory rather than the inverse. Having fast expensive caches near the processor is an attempt to get the memory closer to the proc. What has been happening of late is that lots and lots of on-chip transistors have been spent on the cache. The Cell architecture is a step in the other direction. They want to spend those transistors on processors instead of memory.

At the limit of this idea you would see something like a super-granulated architecture with a processor on each memory chip. Imagine a PC with 32/64/whatever cell processors *and* no classic "processor socket" on the motherboard, just some DIMM-like "cell" slots. Each proc would have exclusive access to the memory on its own chip and all would communicate via some sort of bus or fabric of links. So, instead of one mega proc with tens of millions of transistors(perhaps half would be cache) at 4GHz with a 400MHz x 32, 64, 128, whatever bit width memory bus you'd have maybe 64, 128, 256? simple ARM-like procs at 400MHz each with something like 400MBs or more available memory bandwidth per proc.

Of course the extreme limit would be to have millions of 1 bit processors, but I don't think that anyone is proposing that just yet. Things do get more and more neuron-like as you approach this limit, interesting eh?

Re:Think another way

[CaptKilljoy]

>...ARM-like procs at 400MHz each with something like 400MBs or more available memory bandwidth per proc.

This could have been done already with ordinary procs; it hasn't because once you hit an area that's serialized, your app winds up effectively running at 400MHz on a single proc. (More of a problem for desktop applications rather than server applications.)

>Of course the extreme limit would be to have millions of 1 bit processors, but I don't think that anyone is proposing that just yet.

How about 64k worth in the form of the Thinking Machine's Connection Machine CM-1?

Other Things to Consider...

[DarkGeek]

Besides the real world hardware issues that will likely come up, the author of the article didn't really cover the "politics" that will be involved in trying to make Cell replace x86 processors.

Most companies, unless some radically new programming techniques are made, are going to have a hard time switching from traditional software programming to one based on the Cell's parallel computing. Even he said it's a pain programming software that will correctly execute in a parallel environment. The only ones likely to switch over with any swiftness will be the game and high end computer industry. But I doubt someone would need a 30 GHz processor to use MS Word or other such programs. That's where most of the commercial PC business is located, using "work" programs that don't require blazing fast processors. And even with most of the computers being in the business sector, Cell is unlikely to (initially if even for a while) make a dent in the server and database markets.

Switching back to the hardware front, what about all the data being fed to the processor to compute? It comes straight from storage devices, whether they be flash, optical, magnetic, whatever. The data transfer rates for these are not close to utilizing the full potential of the Cell processor and won't be for a while. On top of that, and being my specialty, because Cell can handle so much so fast, they will require huge storage mediums along with huge data transfer rates. This will be needed due to the sheer size of new software written for Cell, which will be quite large simply because Cell can handle it quickly. Terabyte/in^2 storage space, what I am working on using nanoparticles and what will be needed for Cell applications, is at least several years away and will cost quite a bit when they come out. Most people won't want to spring for these hard drives or other storage mediums needed to fully use Cell. So until then, Cell might actually be more than we need or can handle. It's going to require leaps and bounds in other areas of the computer industry and that may be a little harder to do.

But then again, these huge leaps and bounds may come about tomorrow so who know what will happen. Science and Technology are advancing so rapidly that its anybody's guess as to what will happen. Sometimes all it takes to rule the world is just being at the right place at the right time with the right tools. Just ask Bill Gates.

Re:multicore, stream-processing, vector-oriented B

[Anonymous Coward]

quote: No cache for CPUs? A breakthrough? Hello! Both PSone and PS2 have the so-called scratchpad, which is what the Cell seems to have: a cache which has to be managed explicitly by the programmer. Breaking news: This is a royal pain in the ass. And calculating bandwidth when reading from this tiny scratchpads makes about as much sense as calculating the speed at which a x86 processor can execute MOV EAX, EBX.

sheeesh... If you would stop for a moment to think about it, you'd notice that this architecture is vastly different from your average PS2 arch. The "caches" are actually one per APU which means that these APUs have a kind of access to "locked-down" cache memory - this is basically what most modern signal-processing CPUs do. Take a look at some DSP code specifically tailored for the PPC440 cpu, it explicitly takes advantage of the ability to lock the CPU's caches. I'm sure the main processing unit WILL have L2 cache, though.

William Dally's Streaming processor. Theft ?

[zymano]

Parallel processing is nothing new but this chip sounds like Professor Dally's custom stream processor, the Merrimac He lectured congress on the need of a vector/streaming processor supercomputer because the current supercomputers are inefficient. His Stanford website. Description of cell sounds just like his processor. Even the drawings. The only differene is the location of memory. He makes the point that memory should accessed fast and installed on the chip. That's whats different between the two.

Now *that's* interesting

[GPS Pilot]

From TFA,

The Cell could be Apple's nemesis or their saviour, they are the obvious candidate company to use the Cell. It's perfect for them as it will accelerate all the applications their primary customer base uses and whatever core it uses the the PU will be PowerPC compatible...

The Core Image technology due to appear in OS X "Tiger" already uses GPUs (Graphics Processor Units) for things other than 3D computations and this same technology could be retargeted at the Cell's APUs. Perhaps that's why it was there in the first place...

If other companies use Cell to produce computers there is no obvious consumer OS to use, with OS X Apple have - for the second time - the chance to become the new Microsoft...

PC manufacturers don't really care which components they use or OS they run, they just want to sell PCs. If Apple was to "think different" on OS X licensing and get hardware manufacturers using Cells perhaps they could turn Microsoft's clone army against their masters. I'm sure many companies would be only too happy to get released from Microsoft's iron grip. This is especially so if Apple was to undercut them, which they could do easily given the 400% + margins Microsoft makes on their OS.

Licensing OS X wouldn't necessarily destroy Apple's hardware business, there'll always be a market for cooler high end systems [Alien]...

A Cell based system running OS X could be nearly as cheap (depending on the price Apple want to charge for OS X) but with Cell's sheer power it will exceed the power of even the most powerful PCs. This system could sell like hot cakes and if it's sufficiently low cost it could be used to sell into the low cost markets which PC makers are now beginning to exploit. There is a huge opportunity for Apple here, I think they'll be stark raving mad not to take it - because if they don't someone else will - Microsoft already have PowerPC experience with the Xbox2 OS...

Cell will have a performance advantage over the PC and will be able to use the PC's advantages as well. With Apple's help it could also run what is arguably the best OS on the market today, at a low price point. The new Mac mini already looks like it's going to sell like hot cakes, imagine what it could do equipped with a Cell...

That was X-Box

[Augusto]

> shots from the N64 were CG renders that still aren't matched in the most powerful PC or console games today.

X-Box is the one that was caught making up screenshots and passing them as real time. We found out on a message board about one of these pictures, and the official X-Box site had to clarify that those were indeed pre-rendered shots (the game was Amped).

http://sellmic.com/index_old.html

Re:That was X-Box

[DeadScreenSky]

Actually the N64 was the most famous console that passed off prerendered CG as ingame graphics. The unreleased Robotech: Crystal Dreams is a great example of this widespread practice. You can find some early screenshots here. Some of those 'gameplay shots' are pretty nice looking, even by current game standards! What the game actually looked like can be seen here. It's especially funny because these graphics might have been possible on the PSX or Saturn as prerendered movies - the cartridges of the N64 really didn't allow that though. You simply couldn't possibly get that kind of graphics out of the N64.

(Nintendo even released two arcade games before the console launch, supposedly using the actual "Ultra 64" hardware. The real N64 was never able to even get close to the graphics they featured...)

The Amped1 screenshots at least approximated what the final game looked like. You had real environments, textures, etc. Just some fake concept art models and lensflares were added. The real game actually ended up looking better... (And Amped2 just destroyed both the concept art and Amped1!)

yawn..

[bored]

This kind of crap comes out every year or two, the x86 is the target for everything to beat, or at least in the eyes of the /. crowd. For some reason people have these biased ideas about the x86 sucking, or some particular arch with a small benefit in one area or another being able to leverage that benefit to completely take over a whole market. Lets see, x86 killers.. the itanium, the ppc, the cellphone (mostly ARM), java, the PS2, going back further the whole group of chips labeled as "RISC" even though after a generation or two non of the RISC chips were RISC. etc...

The x86 is popular because of a number of reasons, each of which may be more important for a particular customer base. Its more a function of being a generalist. It tends to be "good enough" for most general purpose computing needs and its cheap enough, and well understood. Just because some arch comes along that is lower power, faster, higher bandwidth etc.. doesn't mean that it will be able to leverage that one advantage enough to upset the massive installed base of x86's.

Anyway, in this case, having to rewrite all your OS and software doesn't bode well for the Cell processors in becomming the x86 killer, at least not for the next 5-10 years or so.

Re:You're falling for the hype- hook, line, sinker

[kai.chan]

You can tell what the real-world performance of an unreleased chip is going to be, just by looking at a simple top-level diagram?

So, instead of providing reasons why my claims are false, you go on to question valid points with sarcasm without providing a single bit of evidence to support your claim? I am not the one not talking out of my mouth.

The write-up is not PR. It is written by someone who figured out the architecture from a patent. Tell me, what processor on the market today has a main processor and 8 arithmetic units attached in a parallel fahsion. In a home system, most computing power goes into audio, and 2D and 3D visualization, much of which are parallelizable. Tell me which current Intel and AMD processor has the capability of pipelining a video encoding/decoding the same way that Cell does. Tell me which Intel and AMD processor today can produce results for as many vector problems as the Cell can.

*ahem*

[Frobozz0]

IBM, you know-- the people that make RISC chips-- would tend to disagree with you:

http://www-306.ibm.com/chips/techlib/techlib.nsf /p roducts/PowerPC_970_and_970FX_Microprocessors

Oh, and Google:

http://www.google.com/search?hl=en&lr=&oi=defmor e& q=define:RISC

Taking a definition of a word from 10 to 15 years ago and applying it to today's language is, well, ignoring the developments of the past decade. The term RISC simply means that the instructions used are fewer and less complex. It also means that design goals are to be most efficient per clock cycle, as opposed to brute force. In addition, they require fewer transistors, which makes them run cooler and be cheaper to produce. It's not the narrow definition you make it out to be.

I suggest you buy the G5 you've been having dreams about and finally address that itch.

Is this a "system on a chip"?

[ThreeGigs]

This really seems like the proverbial system on a chip. If you don't need a PCI bus, or AGP bus, or any of the equivalents, the system chipset becomes a rather small bit of silicon. AMD has shown us how to put the memory controller on the CPU die. So why not go the extra step and combine what's left of the chipset with the CPU? And then, just to make things interesting, why not take the computational units out of the graphics card, and put them where they can be used more easily. Consider shader 4.0 Apps using the video card as another (rather specialized) processor. But instead of putting all those vector units in a video subsystem, they've been put on the CPU where *any* program could make better use of them.

What this seems like to me is simply that [TSI] has taken all of the processing power which would have been spread across several chips, and put it all on *one* chip. With a memory controller and a stripped down system chipset. A modern PC is simply a bunch of coprocessors all in one box. It makes sense to take all those separate coprocessors, put them on one chip, and standardize the programming interface. It's like being able to use the processing power of the dsp on your sound card to make your graphics a bit faster.

I think the "GPU" nVidia is designing probably handles little more than the video buffer and a few generalized, non computationally intensive duties.

Oh, and 85 degrees celcius with a heatsink is *excellent* thermal performance. If you don't think so, try running your processor with just a heatsink, fanless.

Simple

[westyx]

Magic. IBM (with the selloff of their harddrive division) now has no need to grind up pixies for the magic "pixie" dust they use to make harddrives. So they have all these pixies and nothing to use them in - so they'll be used to cool the chips :)

IBM said Sell not Cell.

[Pinback]

It is all a big misundertanding between Sony and IBM.

IBM told Sony it was going to "Sell" its PC busines. Sony has been telling everyone about IBM's "Cell" PC ever since.

Seriously though: For all we know, the PS3 may have four cells. (One CPU core, and three "APU" cells.) One APU for the boobs, one APU for animated low polygon count "hair", and one for inane dialog.

Maybe the new splice() based pipes in Linux can be used to move data between APUs.

Ever notice how much in common the Gamecube and the Mac Mini have in common?

Seeing DRM in Cells?

[MonkeyBoyo]

What nobody has objected to is the IBM claim that the architecture has:

On-chip hardware in support of security system for intellectual property protection.

Much of the curent discussion has been on how to program and coordinate all the little digital signal processors (DSPs - aka DPUs). I think these questions are moot because the envisioned DRM (digital rights management) will make the "cell data" and "cell programs" uninspectable. Even by the the on-chip PUs (processor unit - something like a power-Mac running Linux).

In other words, media data and processing algorithms will be behind an impenetrable DRM hardware wall. "Cell programs" (the little vectorizable data manipulators) will be trade secretes. Outsiders that want to program something new will only be able to string together DRM approved cells. For example, there might be an approved MPG6 cell that will report meta-data found initially in a MPG6 stream but Rights Management interests will never permit any cell that exports all of the MPG6 data.

Why does the recommended single chip PE (processing element) include 8 DPUs? My guess is that a certified library of Cell Programs will not allow anything to be sent off chip that is not strongly encrypted. Thus one might have an 8 DPU chip where 3 are used to decrypt the input, 2 to do the actual processing, and 3 are used to encrypt the output. This off-chip disadvantage is a strong reason for putting multiple PUs and their 8 DPUs on one chip - If intercommunication between Cells cannot be detected externally then there is no need for the encryption/description stuff.

Re:Seeing DRM in Cells?

[Fallen_Knight]

And i think that alone will prevent cells from takeing over the PC market... Or being used for anything aside from embedded deviced and consoles.

People get mad when they can no longer do something they used to (or want to). And alot of people are VERY used to being able to rip music games and video, or hell even create video, tinker, hardware/software hack, whatever.

The current copy protection most likely doesn't get people up in arms because theres nothign that can't be beaten yet.. any game can be copied, and music is even easier. And if you dont' know hwo to copy it someone does and its on the net for you to download.

Re:Seeing DRM in Cells?

[Mindmatter]

I doubt this will be an issue, generally software needs to be programed to use the DRM in the first place. I doubt there will be as much encrypting and decrypting being used, if any at all in the way this poster mentioned.

It's the architecture

[Mr2001]

PC software is written around certain assumptions, for example: moving data from one place to another is slow, so you load everything into RAM before you need to work with it, then load all your graphics data into video RAM before you need to display it, etc.

The PS2, and I suspect most other consoles, doesn't work that way. It has a few processors running in parallel, and MASSIVE bandwidth, and you don't need a lot of temporary storage because it's easy to just pass data from one chip to another as you work with it.

Porting PC software to the PS2 without dramatically restructuring it will result in obvious problems, since the PC software relies on a lot of temporary storage. Porting PS2 software to the PC will also result in problems, not because the PC is a poor design, but because it's just different from the PS2.

Re:not a new architecture, and it's going to be to

[alan_dershowitz]

This contributed to the lack of developer support for the full feature set of the saturn. Scant few games actually took advantage of the dual CPUs.

Re:You're falling for the hype- hook, line, sinker

[be-fan]

Okay, one by one:

The CS monitor says it exactly like it is: PIII gets 2 gigaflops, and the PS2 get's 6.2 gigaflops. Consider the former is just for the CPU, and the latter is for the whole system (CPU + GPU), this isn't outrageous. The "twice as fast as a supercomputer" stuff was editorializing on the part of the writer.

In the Time article, Sony doesn't exaggerate at all. The PS2 really can sustain ~70M polygons per second when they are simply shaded. The "microprocessor as powerful as a supercomputer" stuff is again, editorializing on the part of the writer.

Show me actual technical literature that shows Sony exaggerating about the PS2. They take favorable numbers (eg: flat-shaded polygons instead of textured polygons), but they are for the most part verifiable. Note that the presentations about Cell aren't the same sort of thing as fluffed-up magazine articles. They are professional presentations, stuff written by geeks for people who know what a DMA controller is. If they are making stuff up for this, they'll be discredited by their peers. I doubt Sony and IBM engineers would take those risks.

Ummmm...

[TR0GD0RtheBURNiNAT0R]

If you've really got an athlonFX and an X850, your comp should be making mincemeat out of consoles in every department. (Although thought #3 is correct; the OS in a console is very simplistic.)

My 2 year old athlonXP 2000+ with a 9600 pro and 256MB of RAM runs NFSU2 much prettier than an xbox/PS2 can, with a perfectly playable framerate, all while playing MP3s in the background, and running that big, bloated monstrosity known as windows.

This is based on an old patent, not the real Cell

[Anonymous Coward]

Boy did this guy sucker you in. Cell is similar to this concept, but changed from the early patent filings.

Re:You're falling for the hype- hook, line, sinker

[Anonymous Coward]

> They take favorable numbers (eg: flat-shaded polygons instead of textured polygons), but they are for the most part verifiable

It's worth noting that a whole lot of Japanese games use only shaded polygons. Even now, most console games use less textures than a PC game, and use algorithmic tricks instead. It took developers years to get their heads around the PS2's bizarre VPU design, but it turned out to be damn powerful in the end (it manages to stay reasonably competitive with the X-Box even though it's a full generation older and has half the RAM).

Hopefully the API will have a better translation job done this time. CLITIndex anyone?

Just one small detail you omitted

[Anonymous Coward]

Windows NT was written as the successor to DEC's VMS operating system. That's why the MIPS base. Add 1 to (each of) VMS, and you get WNT.

Bit of history for the kids.

Re:You're falling for the hype- hook, line, sinker

[Anonymous Coward]

Cell doesnt do any of that. Cell is at this stage a few patents and a dozen or so press releases.

Tell me which current press release has the capability of pipelining a video encoding/decoding the same way that an intel/amd processor does?

Re:How about quotes from the story submission?

[Anonymous Coward]

The fact that 99% of consumer applications other than games dont require additional CPU power.

Your typing...

[Anonymous Coward]

hurts my eyes.

Re:Nicholas Blachford is an idiot. Please don't re

[Eric604]

I predict the next step in software engineering for power hungry applications will be to move away from the OO model and instead design data processing components that execute in parallel. It would be something like a hybrid between VHDL and C++.

Um,

[zak]

That's FreeBSD, bubba. And NeXT started off on 68K, not x86 - they got to x86 late in their lifecycle.

Re:Um,

[EvilTwinSkippy]

I'm not worthy! (Begins ritual self mutilation)

Re:multicore, stream-processing, vector-oriented B

[Anonymous Coward]

Maybe you should of read the article instead of trying to impress us with your rad knowledge oh master. The Linux kernel and GCC will be compiled, no doubt about it.

Re:Nicholas Blachford is an idiot. Please don't re

[Anonymous Coward]

I predict the next step in software engineering for power hungry applications will be to move away from the OO model and instead design data processing components that execute in parallel.

Data processing components that execute in parallel...that perhaps communicate by passing messages? Congratulations, you have just defined OO programming.

Re:You're falling for the hype- hook, line, sinker

[i41Overlord]

The write-up is not PR. It is written by someone who figured out the architecture from a patent. Tell me, what processor on the market today has a main processor and 8 arithmetic units attached in a parallel fahsion. In a home system, most computing power goes into audio, and 2D and 3D visualization, much of which are parallelizable. Tell me which current Intel and AMD processor has the capability of pipelining a video encoding/decoding the same way that Cell does. Tell me which Intel and AMD processor today can produce results for as many vector problems as the Cell can.

The mistake you are making is that you're comparing a processor that only exists on paper and press releases to an actual, physically present processor that has already been designed, tested, marketed, and used by millions of people.

I could write a technical writeup for an imaginary processor that has xxxxx terraflops, only uses 1 watt of electricity, and only costs $5 to make. It's easy to do that stuff on paper because you don't have to deal with the technical hurdles and limitations that reality imposes. Talk is cheap. It's another thing entirely to actually produce the product and have it do all the things you said it would.

What if in the prototype stage they discovered that the working silicon doesn't perform as the design was expected to? What if they can't ramp up the clock speed on it due to unexpected reasons? What if they can make it but it ends up costing them more to produce than anticipated? These are things you have to work out before you can sell an actual product. But a write-up doesn't have to deal with these problems.

That's why I say that talk is cheap. It's easier to talk about something than to do it. Wait until the cell processor does come out and see how it performs. Promising the stars is very easy. Delivering the goods isn't quite so easy...