Slashdot Mirror
Intel Pledges 80 Core Processor in 5 Years
ZonkerWilliam writes "Intel has developed an 80 core processor with claims 'that can perform a trillion floating point operations per second.'" From the article: "CEO Paul Otellini held up a silicon wafer with the prototype chips before several thousand attendees at the Intel Developer Forum here on Tuesday. The chips are capable of exchanging data at a terabyte a second, Otellini said during a keynote speech. The company hopes to have these chips ready for commercial production within a five-year window."
...Imagine a Beowolf cluster of those!
(Runs in shame.)
'Sensible' is a curse word.
But it still can't tell me 1/0.
Exchanging data (data transfer) is not the same thing as operations per second. The post seems to either be confusing the two or stating that the chip does both. I guess I need to go read the article now and find out...
This will finally run Vista, right??? Maybe? Hopefully?
Unto the upright there arises light in the darkness...
Unfortunately, they'll all choke on a shared memory bus :-)
What massively parallel tasks would possibly need 80 cores? I can see uses for two, maybe 4 cores but what are advantages of 80 core chip as opposed to system with 40 2-core processors we can have now?
promised us 8-10Ghz Pentium4 CPUs when they started with the P4 "Willamette"? Or how they promised us 5GHz Prescotts?
I'll rather wait and see what I can actually buy in 5 years. No need to trust a vendor so far in the future what they can do.
Faster processors are great, but when will we see massive improvements in data storage...
Imagine the pain of having to write a functional applications with so many cores. I hope the interconnect will be very very fast. Otherwise writing massively scalable parallel algorithms will be masssively painful. And with so many cores, one will need multiple independants memory banks with some kind of NUMA. And writing apps for those things isn't fun. You have to spend so much time caring about the parallel stuff instead of caring about the problem.
That's hot!
(/ducks)
I seriously hope that power consumption and heat disipation are really attacked before these things come out. Can you imagine needing a 200-amp service and liquid nitrogen cooling for something like that right now?
This is hilarious, because if this goes out on the market there's not going to be many operating systems capable of scheduling on that many chips usefully. OS X can't do it, Windows can't do it, and nor can BSD. But Linux has been scheduling on systems with up to 1,024 processors already :)
With the heavily threaded nature of BeOS, even demanding apps would really fly on the quad+ core cpus that are preparing to take over the world.
Not that you couldn't do threading right in Windows, OS X, or Linux. But BeOS made it practically mandatory: each window was a new thread, as well as an application-level thread. Plus any others you wanted to create. So to make a crappy application that locks up when it is trying to do something (like update the state of 500+ nodes in a list; ARD3 I'm looking at you) actually took skill and dedication. The default state tended to be applications that wouldn't lockup while they worked, which is really nice.
Slashdot Patriotism: We Support our Dupes!
A chip with 80 Floating Point Units is not the same as a chip with 80 Cores.
Doesnt moores law state speed will double every 18 months? Lets see. We have 2 cores now. 4 in 18 months, 8 in 36 months, 16 in 52 months, 32 in 68 months.....wait a min. 5 years is 60 months. Are they going to go faster than moores law. I doubt it.
Lets wait and see, but I think this is just an empty promise.
In other news, Gillette pledges a razor with 81 micro blades. 80 blades are individually controlled via Intel's new 80 core processor. The 81st blade is available just because..
Just in time for Vista!
Seems like (classical) microprocessor computing is quickly converging with reconfigurable computing. In the embedded systems world there has been a recent emergence of something OTHER than a re-hash of FPGAs. FPOAs seem to be the most prominent one and approach reconfigurable computing through a heterogenous array of ASICs which come in three flavors : Multiply accumulators, general purpose ALUs and speedy register files (which can be configured as a FIFO,RAM or sequental read random write memory block).
..oh wait that's already happened with embedded powerPCs in Xilinx's Virtex 2 and Virtex 4's.
Quite a departure from the relatively homogenous composition of an FPGA which is basically a bunch of SRAMs connected with each other and a ton of switches.
Pretty soon we should be seeing FPGAs acting as co-processors..oh wait..that's already possible (see XD1000 which has an Opteron and an Altera Stratix(?) communicating via hypertransport). Or maybe we'll start seeing micrprocessors acting as a periphereal to an FPGA
Wow, good point. I bet Intel never once stopped to think about THAT.
I sincerely doubt this will make it anywhere near Fry's or CompUSA, assuming it launches in +5 years. Most likely academic, corporate (think of the old days and mainframe number crunchers on wallstreet), and scientific.
Simply cheap teraflops for custom applications.
Of course, everyone thought it was a great idea when Cell announced they could do 64 or more cores. But since this is
https://www.accountkiller.com/removal-requested
Ok. We now know Vista will be 5 years late. :)
Charles Wyble System Engineer
The costs to make use of 80 cores (you're going to need hugely complex chips and hugely complex memory buses) mean that these chips will be severe overkill for PCs and will be outside any typical user's price range. They're only going to be useful for a a few servers in very niche applications. If there's only demand for, say, 10,000 of these chips in the world then they're going to be extremely expensive.
I smell marketing horseshit. I think they're just saying this to get people to start thinking of multi-core options. Most people don't see the need for multi-core (even 2 core) systems. By saying you'll get 80 cores in 5 years makes people start thinking that they should start using 2 or 4 cores now.
Engineering is the art of compromise.
Today, a 2 CPU x 2core computer can actually be slower than a 2x1 or 1x2 core for certain "cherry picked to be hard" operations due to the OS making incorrect assumptions about things like shared/unshared cache - 2 cores on the same chip may share cache, two on different chips may not - and other issues related to the fact that not all cores are equal as seen by the other cores.
In an 80-core environment, there will likely be inequalities due to chip-real-estate issues and other considerations. The question is, will these impacts be felt at the code level, or will the chips be designed to make these differences invisible? If the former, will the OSes be designed to use the cores efficiently, or will they simply see "80 cores" and, out of ignorance, make poor decisions when allocating tasks to various cores? If the latter, what performance penalty will be incurred?
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
I can just hear Microsoft gloating over this "Wow! Now we can add all the DRM our real customers want without any apparent performance penalty."
"Quick, run out and buy stock in every power and cooling company you can!".
Of course the malware will be able to fly faster on the new Microsft DRM 2010 Media Center release (due in 2012)
It will take another 15-20 years for software to catch up.
512 cores should be enough for anyone.
Gosh, do these guys *ever* learn? I still fill the pain from the broken promise on the 10GHz P4.
We had to have that 10GHz Pentium somewhere around this time I think. Reality? We don't even have Pentium anymore, after they redesigned it.
Intel wants to wow everybody with flashy predictions about their own future, but Intel, but what people care about is whaat they are selling *now*.
They seemed on the right path with Core 1/2, but it seems they are back to the silly "I have more of feature X" marketing.
An 80-core chip with RAM attached directly to the processor chip, as TFA discusses, is going to have an advantage in transferring data between cores, and plus it'll probably be a lot smaller. Than 40 dual core (or 20 quad core) chips.
It would quite likely be cheaper and consume less power.
"Our morality is good, theirs is repressive."- Partisanship Rule #3
That's the sound of 5 million geeks ejaculating simultaneously...
I'll form my OWN solar system! With blackjack! And hookers!
It's not the hardware that is behind the times it's the god damn software. -VoG-
640 cores ought to be enough for everybody
Slashdot, fix your code or at least hire someone who is competent at it to do it for you.
It will take another 15-20 years for software to catch up.
That's what every hardware engineer says when CPUs progress from being way, way too slow to way too slow.
Are you happy with the speed of your computer? I'm not. I'm still waiting for the hardware to catch up with the demands of last year's software.
I trust Intel in everything they promise.
They are great, that's why I'm posting this on a 5.0 Ghz Pentium 4 processor and... oh wait
Look here for more information on the technical specs; http://www.eetimes.com/news/latest/showArticle.jht ml;jsessionid=13DOWA104O1JYQSNDLSCKHA?articleID=19 3005741
From the article:
"When combined with our recent breakthroughs in silicon photonics, these experimental chips address the three major requirements for terascale computing -- teraOPS of performance, terabytes-per-second of memory bandwidth, and terabits-per-second of I/O capacity," Rattner said in a statement. "While any commercial application of these technologies is years away, it is an exciting first step in bringing tera-scale performance to PCs and servers."
I'm still in the "believe it when I see it" phase.
Scheduling isn't a one size fits all process. What works at 4 cores doesn't work at 40 and so on. As for other operating systems, FreeBSD has been working quite actively on getting Niagras working well with their sparc64 port. I've been saying it didn't make sense until this announcement. I figured we'd have no more than 8 cores in 5 years. We'll see what really happens.
The BSD projects, Apple and Microsoft have five years. Microsoft announced awhile back they want to work on supercomputing versions of windows. Perhaps they will have something by then. Apple and Intel are bed partners now. I'm sure intel will help them.
What this announcement really means is that computer programmers must learn how to break up problems more effectively to take advantage of threading. Computer science programs need to start teaching this shit. A quick you can do it, go get a master's degree to learn more isn't going to cut it anymore. There's no going back now.
MidnightBSD: The BSD for Everyone
Slashdot is powered by your submissions, so send in your scoop^H^H^H^H^H ripped off post from someone else's blog.
Isn't it kind of immature for a company to think "hmm, we have had relative sucess with dual core processors over our competitors. . so lets fit as many cores into a mobo as possible . . .that will get amd" If I recall properly, didn't the same thing happen with resistors on circuit boards back when radios and such were new? People just kept trying to fit more and more on a board rather then researching better technology. While I believe that multi-core technology needs to be developed further, there are also other things for intel to be researching.
Klingon Software is not released, it escapes, inflicting terrible damage onto the enemy as it does
From TFA:
Intel's prototype uses 80 floating-point cores, each running at 3.16GHz, said Justin Rattner, Intel's chief technology officer
So it's some sort of Pentium-ish beastie with 80 floating point units, not an 80 core CPU.
Although, I could easily find a use for either one. Just off the top of my head, an 80 FPU machine would be an excellent science/simulation machine. And it'd probably make some fairly decent graphics for games. You could use that much floating point for voice recognition without taxing your machine very much. You wouldn't need a special graphics card anymore either - just allocate a dozen or so FPU units to plotting 3D graphics and simply stamping them into a display memory buffer.
An 80 core machine would be excellent for system simulation. Imagine things like Wine and VMware running with that much elbow room. Or you could split the taskload up for networked applications like WETA's render lab. The whole render lab would fit on a desktop.
More computing power is always better. Well, until The Terminator finally makes it back to Intel's lab and smashes the prototype, anyways.
Weaselmancer
rediculous.
Things like memory bandwidth are already constraining 2-core chips. The only way to effectively mitigate this is to make wider bus paths. That's relatively easy for 2 core chips, but to get any benefit from 80-core chips you're going to need 40x the memory bandwidth you have now. That means huge pin-outs, huge amounts of RAM, huge everything.
These are not going to be systems that every college department can afford.
Sure people won't care if the cores are there, but they will care about the price if that impacts on the whole system cost.
Engineering is the art of compromise.
Really if your read the story it is 80 floating point cores! It would be be ideal for many graphics, simulation, and general DSP jobs.
What it isn't is 80 CPU cores.
Really interesting research but not likely to show up in your PC anytime soon.
With all these multi core chips I am waiting for more specialized cores to start being included on the die.
After all a standard Notebook will have a CPU, GPU and usually a DSP for WiFI. Seems like linking them all with something like AMDs Hyper transport could offer some real performance gains. Imagine a core optimised for TCP/IP to go along with the DSPs audio and wifi and your GPU all closely tied to together with Hyper-Transport links.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
... not for the high price Gasse wanted for it, but for what 3COM got it for. They need that pervasive multi-threading now more than ever. NEXT was good and all, but are they really going to be able to backwardly refine the whole bit? Oh well, at least they've got plenty of old BeOS employees. The pervasive beach-balls however make me wonder what they're doing all day, new kernel?
Cwm, fjord-bank glyphs vext quiz
...in search of a problem. But don't you worry, those problems will come along soon.
And AMD recently promised the development of an 81 core processor...
Register the editry.
not 80 general purpose integer cores. They're essentially copying the Cell design with large numbers of DSPs each of which has a local store RAM burned onto the main chip. Is this a good idea? Guess we'll find out with the Cell. What interests me most about this announcement is not the computing potential from such a strategy, but that it's an obvious response to IBM and Sony technology.
You fools! Do you have any clue how much Oracle licenses will cost for this thing?
Can you imagine the heat sink and the fan?
You could peal the paint off the cubicle with that kind of heat and wind
-- I am the NRA, enough said...
A couple of things to mention here. Many years ago I read an Intel road map for the x86 processors. It was more than 10 years ago, less than 20 I think. In it they said they would have massively multicore processors coming along around now. They may have forgotten that and reinvented the goal along the way, companies do that. But, they really have been predicting this for a very long time.
:-) The counter example everyone came up with was, "well, if that is true why would AMD buy ATI?" The answer to that is simple, they want their patent portfolio and their name. In the short term it even makes sense to put a GPU and some shaders on a chip along with a few cores. At the point you can put 16 or so cores on a chip you won't have much use for a GPU.
The other thing is that with that many cores and all the SIMD and graphics instructions that are built into current processors it looks to me like the obvious reason to have 80 cores is to get rid of graphics coprocessors. You do no need a GPU and a bunch of shaders if you can throw 60 processors at the job. You do need a really good bus, but hey, not much of a problem compared to getting 80 cores working on one chip.
With that kind of computer power you can throw a core at any thing you currently use a special chip for. You can get rid of sound cards, network cards, graphics cards... all you need is lots of cores, lots of RAM, a fast interconnect, and some interface logic. Everything else is just a waste of silicon.
History has shown that general purpose processing always wins in the end.
I was talking to some folks about this just last Saturday. They didn't beleive me. I don't expect y'all to believe me either.
Stonewolf
This is the last 3 years of Intel, all over again. Only now the megahertz race is replaced with the multi-core race.
Intel will create the "CoreScale" technology and make 4, then 8, then 16 cores and up while their competitors are increasing operations per clock cycle per watt per core. Consumers won't know any better, so they will buy the Intel 64-core processor that runs hotter and slower than the cheaper clone chip that has only 8 cores. Then when Intel starts runs up against a wall and gets their butt-kicked they will revert to the original Core 2 Duo design and start competing again.
Oh, and I predict that AMD will release a new rating called the "core plus rating" so their CPUs will be an Athlon Core 50+ meaning it has the equivalent of 50 cores. Queue n00bs who realize they have only 8 cores and complain.
And to think I didn't like history in school. Maybe I just hadn't seen enough of it to understand.
Maybe I'm missing something, but it seems from the article that this is a processor with 80 floating-point units connected together rather than a functional x86 processor.
It sounds more along the line of a Cell processor competition rather than a successor to current x86 processors. With many parallel FPUs, it might be nice for graphics processing and scientific calculations (which should be easier to saturate all those cores), but for everyday computing, would it be very useful?
And perhaps, as the word "core" is gaining attention at the marketing department and the general public, the definition of a core is becoming more ambigious...
I'll be called "Core Lots Quad-Duodeca"
Rome taught me patience and assiduous application to detail. Virtues which temper the boldness of great, general views.
Assuming cpu is the bottleneck, 1 core per process is a great ideal.
Unfortunately, as others have already pointed out, there are other factors, such as memory bandwidth, etc., that will make an 80-general-purpose-core single-cpu machine far slower than a well-designed, 80-cpu machine or even 80 single machines, at least for independent tasks that contend for the same non-cpu resources.
80-core systems will shine where there is a lot of inter-thread/process communications and where there is little contention for resources other than the CPU.
Besides, it looks like this isn't a real 80-general-purpose-core machine anyways.
Knowledge is how to play a game, intelligence is how to win, wisdom is knowing what game to play.
Duke Nukem Forever is just going to fly on one of those.
Um, OS X already can. They just don't sell the boxes yet.
p =6
http://www.anandtech.com/mac/showdoc.aspx?i=2832&
There are two types of people in the world: Those who crave closure
Software hasn't really improved for maaany years now, Spreadsheets and Word Processors are more colourful, higher resolution. But are these products smarter, better at all? Would a postgraduate write a better doctoral thesis with Office 2007 than with - say - Word 6.0? Is image manipulation thaat much better with the latest photoshop than with PS 5.5? With some minor exceptions the answer is clearly no.
- We were promised Virtual Reality with VR Helmets more than 10 years ago - is this _just_ a matter of hardware?
- Smart voice recognition? Anyone tried it lately? Anyone tried to write pretty standard letters with it? Desastrous.
- Intelligent assistents, understanding the user's needs? Operating system/application wizards that improve it's capabilties while you're working with 'em?
The applications are missing, they're faster, more colourful, higher resolution, antialiased... but still DUMB.
Computers are already pretty powerful, please start and make the software smarter, not faster.
CPU power is not that important anymore.
I can't wait until i have one of these in my cell phone.
i support the right to offend.
Virtualizing your IT server infrastructure into fewer servers with more cores has a lot of benefits. One is less power usage overall, the other is better utilizing the total processing power. Not to mention things like better disaster recovery since the OS is no longer tied to specific hardware.
Sometimes my arms bend back.
Ok, so the Core 2 extreme will have 4 cores, to be followed by a core 2 quad also with 4. Hard on the heels of the Core Duo and the Core 2 duo. Is this just fallout from the 486 generic trademark problem, thus ensuring that their processor names will never make sense again? /me anxiously awaits the Core 2 octo with 16 cores, 4 m's and a silent q.
No one will ever need more than 640k^H^H processors.
Eloi, Eloi, lema sabachtani?
www.fogbound.net
The question is what will be released first Vista or 80 core Intel?
WASHINGTON (Reuters) In lights of Intel's 80 Core Processor pledge in 5 Years, scientists are worried that Richard Branson's pledge is now too little too late.
[alk]
All the chip manufacturers are hitting the Mhz barrier at around 3.xGhz speed. (Sure, some people are overclocking CPUs to 4+Ghz but not in a practical way that a CPU vendor could sell as a mass-marketed product.)
When Intel saw the success of the Core Duo, a light went on over someone's head and they realized "Hey! This is the solution! Keep increasing the number of cores!"
Right now, people against this idea are usually pointing out the lack of software/OS support for it as the reason it's not a good idea. But realistically, this is probably the easiest hurdle to overcome. With proper design, an OS shouldn't care whether it has 4 or 400 cores at its disposal. The same logic will get used to hand off processes to available cores, and applications will gradually be optimized for OS's that exhibt this behavior.
Like most things, it's just a matter of giving it some time to get developed.
But the more pressing issue is bus bandwidth. It won't take long for multiple cores to exceed the bandwidth limit to/from the system RAM - and this is already the "weak spot" on current PCs. (We're doing good to get front-side bus speeds of 50% of a CPU's processor speed right now.)
Furthermore, hard drives are a huge bottleneck too. Solid-state memory drives keep getting billed as the solution here - but again, you're talking devices chewing away at available bandwidth on the bus.
If AMD manages to develop higher Ghz speed CPUs rather than continuing down Intel's "more cores = better" path, they stand a good chance of having the superior price vs. performance CPU again in the not too distant future.
I said scale and schedule usefully, not "run". Big, big difference. Last I heard, OS X Tiger only had 6 kernel locks. You expect that to run usefully on 80 cores? Anandtech even said "We definitely had a difficult time stressing 8 cores in the Mac Pro"...
AMD and Intel are both expecting to ship dual processor quad core configurations by the end of this year, Slahdot recently had an article about AMD doing this, Anandtech has a review of a new dual CPU Apple that had dual core chips that they upgraded and had running two quad core chips.
Based on this somewhere around 20 core systems would not be out of place in 5 years on the professional desktop.
or find out what smalltalk or even prolog is really about...
...someone is already trying to figure out how to put this on a graphics card with a terabyte of memory for their gaming computer.
:)
The big question is how these processors interconnect. Cached shared memory probably won't scale up that high. An SGI study years ago indicated that 20 CPUs was roughly the upper limit before the cache synchronization load became the bottleneck. That number changes somewhat with the hardware technology, but a workable 80-way shared-memory machine seems unlikely.
There are many alternatives to shared memory, and most of them, historically, are duds. The usual idea is to provide some kind of memory copy function between processors. The IBM Cell is the latest incarnation of this idea, but it has a long and disappointing history, going back to the nCube, the BBN Butterfly, and even the ILLIAC IV from the 1960s. Most of these, including the Cell, suffered from not having enough memory per processor.
Historically, shared-memory multiprocessors work, and loosely coupled network based clusters work. But nothing in between has ever been notably successful.
One big problem has typically been that the protection hardware in non-shared-memory multiprocessors hasn't been well worked out. The Infiniband people are starting to think about this. They have a system for setting up one way queues between machines in such a way that appliations can queue data for another machine without going through the OS, yet while retaining memory protection. That's a good idea. It's not well integrated into the CPU architecture, because it's an add-on as an I/O device. But it's a start.
You need two basic facilities in a non-shared memory multiprocessor - the ability to make a synchronous call (like a subroutine call) to another processor, and the ability to queue bulk data in a one-way fashion. (Yes, you can build one from the other, but there's a major performance hit if you do. You need good support for both.) These are the same facilities one has for interprocess communication in operating systems that support it well. (QNX probably leads in this; Minix 3 could get there. If you have to implement this, look at how QNX does it, and learn why it was so slow in Mach.)
Actually, such an architecture with careless multithreading can be a disaster if misused. When you have lots of threads running on one CPU, there isn't much penalty, especially if the OS is wise enough not to do anything that might flush the TLB in between thread-switching. All the necessary locks and whatnot will be in cache.
The problem with adding cores is that the operating system scheduler (naturally) is going to spread those threads out across all those cores or CPUs. At that point, you better damn well have broken your application down correctly, pipelining work and being careful not to depend too much on centralized locks, or else the resulting cache-line ping-pong is the only thing you are going to be doing with your chip.
To that end, putting every interdependent program component into a thread context of its own sounds like a disaster waiting to happen, because that's not how you make useful use of lots of cores.
So, how big will the processor be? The size of a VCR or a VHS cassette? Also, will the heatsink's fan have a jet-like flame shooting out of the top?
No they don't. Right now I'm building a Linux kernel and it is only using approx 35% of the CPU. Why? Because my memory and disk are not fast enough. If I swapped out the CPU and kept everything else the same, it would not go much faster. Sure, with a faster motherboard etc I could get better speed, but that is very difficult to scale to 80 cores
As I said before.... to get 80 cores working properly is going to require huge amounts of memory as well as hugely wide buses out of the chips (say 512 bit-wide buses), huge increases in disk rw speecd etc.
Nobody is going to design 80 core systems unless someone is prepared to buy them and nobody is going to design 80-core chips if nobody can show how to design effective systems with them.
For people wanting to crank SETI etc, it is going to be way cheaper to build a cluster with 20 4-core systems.
Engineering is the art of compromise.
Don't please Intel developers to such a horrendous solution. Intel has never presented any technology successful of NetBurst that would scale on this matter. Perhaps Intel plans on buying SGI if this flops, though SGI developers deserve the award for this contract without any branding from Intel. As always, yesterdays Supercomputer technology should be fab-shrunk down to today's PDA wrist-watch. As well, there is no demand for an 80-core Processor.
Yes. PS CS2 is light years better than 5.5 for the imaging professional.
How?
Well, two things that make my life MUCH easier:
1. The Shadow enhance thingie. You can use it to bring out details in a shadowed area. If used badly, it looks like crap, but it can really make or break a photo. Awesome tool.
2. The Healing Tool. It's kind of like the clone tool, but much more sophisticated. I have used it countless times to "fix" a face.
Those are only 2 minor things that are big BIG pluses for me. There are bigger things, but I'm pressed for time right now. I could go on and on, but seriously: Photoshop CS2 is WAY better than 5.5.
RS
Shoes for Industry. Shoes for the Dead.
So, this thing will come out less than after receiving my shiny new Computer Science diploma. Suddenly much of what I know will become useless, thanks Intel.
So think more like Cell with 80 SPEs. Great for lots of vector processing.
No folly is more costly than the folly of intolerant idealism. - Winston Churchill
Seriously, just wait. Don't upgrade until this sucker comes out, otherwise what ever you buy will be obsolete fast. Lash your PC together with duct tape if you must but do not purchase a new pc for the next five years. It's worth the wait. Ditto for buying software. There is no good reason to upgrade until the 80 core comes out. Keep your money in your pocket.
putting the 'B' in LGBTQ+
Ahh, remember the good old days when Intel would point to AMD and start singing COPYCAT, COPYCAT like a child on the playground?
Now, after the failure of Itanic and Netburst, it looks like they are so short of inspiration that they stole a page from the SUN Niagra processor roadmap. Kudo's to Sun, while they seemed to be really messing up on designing silicon in the last 5 years, at least they had the forsight to call this one.
I thought we were supposed to be moving towards Chaos Computing? Would this be a step in that direction?
Is image manipulation thaat much better with the latest photoshop than with PS 5.5? With some minor exceptions the answer is clearly no.
Hah! I am forced to disagree in the strongest possible terms..
Speaking as a former production artist and current art director, the last couple of generations of graphics software have introduced powerful tools that streamline my workflow in ways I find it hard to even fathom. Ok, let's talk about Illustrator, for example. From 10 -> CS Adobe added in-application 3D rendering of any 2D artwork onto geometric primitives. This is something I used to either have to fake, or take out of the application and into a 3D renderer in order to render simple bottle/can/box packaging proofs. Marketing wants to make a copy change? Make the change to the 2D art and the 3D rendering is updated in real time. Oh, and the new version of InDesign recognizes that the art has been updated and reloads it into the brochure layout. Automatically.
This is just one feature out of literally hundreds. This one alone saves me an hour or two a day. Seriously, there are projects I can take on today that would have been unthinkable 5 years ago. Pre-press for a 700 page illustrated book project has gone from a week of painful, tedious work down to 30 minutes, of which 20 is letting the PDF render. Seriously.
Here's the thing, unless you use a piece of software all day, every day, you're really not in any position to comment on how much it has or hasn't changed.
Photoshop (et. al.) are software for professionals, despite the number of dilettantes out there using them for sprucing up their MySpace page.
m-
You catch enchiladas by picking them up behind the head and holding them underwater until they don't kick anymore -VeGas
.. how much will be hot that chip?
-- Giovanni Daitan Giorgi http://gioorgi.com http://www.siforge.org
When will Intel fix their floating point issue?
Is 1994 soon enough for you?
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
Does it really matter? I mean, look at your current process list. If you're like most people, you'll see a lot of idle threads. Each successive processor is less useful. Two processors is great. Four will probably help. Eighty will only be useful to those rare people who need to run massively parallel algorithms.
The cake is a pie
God I'm so sick of these pie-in-the-sky exaggerations of currently hot technologies as "the future".
All it does is depress the market who sits back, content to wait for "something better", even if it never gets delivered.
Hello? Anyone remember Osborne Computer Corporation? I didn't think so.
Sometimes I wish these buzz-speakers would STFU.
Chas - The one, the only.
THANK GOD!!!
64 cores ought to be enough for anyone.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
Holy Shitballs batman!!!!
My posts are definitive. Reality is frequently inaccurate.
The right path is taking all the stuff that's currently on the motherboard and putting it into the CPU. Including some serious chunk of flash memory and (of course) several gigs of ram. This is because performance is already heavily determined by the communications speed between all these things, and putting them in the same IC would allow multi-giga-hertz communication channels.
e rs_in_20.html
Digital Crusader: Computers in 2020
http://digitalcrusader.ca/archives/2006/02/comput
augment your senses: http://sensebridge.net/
Software hasn't really improved for maaany years now, Spreadsheets and Word Processors are more colourful, higher resolution. But are these products smarter, better at all?
So what you're saying is that microsoft office hasn't improved at all. Surprisingly, there's more software out there than office applications, and many of them have indeed benefitted from more powerful computers. Look at digital audio and digital video applications. Digital audio hasn't really been possible without external dsp (a la Protools) until fairly recently - within the last 5 years or so. Digital video is still at the point that digital audio was at 10 years ago, too - we're not going to see personal computers that can truly handle processing of digital video effects with ease for another few years. So while newer, faster computers might not make your word processing experience much better, they're still pretty exciting in other fields.
free music
In other words, get out your functional languages like Haskell and OCaml and use the side-effect free feature set to develop multi-threaded programs. Or do it the hard way with an OO language.
Fascism trolls keeping me up every night. When I starts a preachin', he HITS ME WITH HIS REICH!
Anyone remember Intel NSP (Native Signal Processing)? The chip for it has finally arrived.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
will it run at 10 Ghz?
tada!
No but they'd sure write a better thesis if they did it in latex. word!? You kidding me!?
Yeah, and once an OS is out that CAN stress an 8 core CPU, everyone will be complaining about the bloated code and feature creep.
I read somewhere (either /. or digg) that if we keep pushing the number of cores on a proc, eventually we'll hit a point where power consumption requires more than your standard 120V outlet. Forget architectural bottlenecks and cooling; when can we no longer just plug our computers into a regular power outlet?
Sun has a 64 thread Niagara II processor coming out in 2007. Assuming they double the thread count every year, that has Sun at 1024 hardware threads when the Intel 80 core chip comes out. Personally I'm backing Sun on this processor throughput game.
Zen tips: Pay attention. Don't take it personally. Believe nothing.
This is hilarious, because if this goes out on the market there's not going to be many operating systems capable of scheduling on that many chips usefully.
Solaris has been for a number of years, and Sun is ahead of the game with multi-core CPUs, with UltraSPARC T1.
But I hate Sun as much as the next Slashdot sheep. They did fire me, after all...
Stick Men
I can't wait to overclock one of these! I plan on cooling it with liquid sodium.
It is by the juice of the coffee bean that thoughts acquire speed, the teeth acquire stains. The stains become a warning
So think more like Cell with 80 SPEs. Great for lots of vector processing.
So, the effect is on-CPU ray-tracing, with leftover power to handle everything else. All the fancy GPUs will go out of business and we'll move to cheap and easy ray-tracing. At least from the people that swear by ray-tracing, once it's doable on CPU, we'll never see a separate GPU again.
Learn to love Alaska
...but given the way Intel keeps underspecifying their memory interfaces, you'll still only ever be able keep one (or maybe two) cores busy when continuously streaming data to and from memory at full bandwidth. Can we please get back to a memory-bandwidth-to-floating-point-performance ratio of about 1 byte per flop?
"My life's work has been to prompt others... and be forgotten." --Cyrano de Bergerac
Smart voice recognition? Anyone tried it lately? Anyone tried to write pretty standard letters with it? Desastrous.
Arstechnica tried it lately, and found it peachy. Perhaps you should try it too.
http://arstechnica.com/reviews/apps/speaking.ars
Windows Compute Cluster Server 2003 has a scheduler that can handle more than 80. It's written as a network scheduler, but seems like it would be adaptable to a single system.
People who think they know everything really piss off those of us that actually do.
Too bad we will never see that on the desktop since Intel thinks we don't need them
Since Intel cannot last 5 years without income, they wither and die, just as Osborne Computer Corp did 23 years earlier in what came to be called the "Osborne Effect".
Those who do not learn from history are doomed to repeat it.
Yeah but it's a solution to a non-problem. Imagine all of your cube mates talking to their computer? I curse at it but that's about it.
Please sign petition to restore sanity to our banking system!!!
http://financialpetition.org/
Raytracing alone isn't the be-all end-all, but if you can do what, say, POV-Ray does (which can augment raytracing with an radiosity-based rendering which can really improve lighting realism), but in realtime at a reasonable framerate with fullscreen coverage, that would be great.
I'm not sure 1 TFLOP is enough to do that and do everything else you need to do for a game, though.
Real purpose: spreading stories about superchips years in advance to kill AMD now. AMD is being ignored to death by the press, and Intel is making sure that AMD can't get a foot in the media door.
Worked well in the past, until they couldn't actually produce a faster product. AMD is still cheaper, but you still can't hear much about their products anymore.
Did you even read your Wiki article? It says pretty clearly that sales did dip initially, but they picked up and brought Osbourne back into the black. What killed Osbourne, according to your link, was some bad management decisions on how to deal with $150,000 worth of surplus hardware, and an unexpected jump in the price of their screens.
If you believe everything you read, you'd better not read. - Japanese proverb
"Prediction: within 10 years, Windows will be a Linux distribution." Me, 7-6-2016
Some applications like multiple cores. CPU intensive tasks on my database servers seem to scale very well across multiple cores. But I think it would all become disk-limited well before 80 cores, though...
In theory, there's no difference between theory and practice; in practice there is.
Title and summary have it (slightly) wrong.
Intel's prototype uses 80 floating-point cores.
Very interesting in itself, but not the same as 80 CPU cores, which is hinted at by summary.
Carbon based humanoid in training.
Comment removed based on user account deletion
Duke Nukem Forever developers announced a complete re-write of the game to take advantage of 80 cores. An inside source revealed that the plan is to have the TV sets in naughty mode show HD content with one core per TV. Our inside source (who asked to remain anonymous) also mentioned that there is a reason the initials of the game are DNF.
Find a job you like and you will never work a day in your life.
I've seen a patch that allows BeOS on a Pentium4 CPU, as well as a patch for BeOS on a AMD Hammer; the only limitation is a 768 MegaByte RAM barrier but to think of it honestly would be 768MB is a good limit for any workstation to cope with; and given, technology could scale the Virtual Memmory Manager onto a high-bandwidth Serial ATA or Serial SCSI bus with a solid-state swap drive composed of DRAM modules (advertised on Slashdot, I can't find).
BeOS is still better than Zeta.
Someone who qualifies the seriousness of their application by the number of threads it uses is severely misguided.
Real applications manage thousands of asyncronous activities without needing thousands of associated threads and task switching overhead. You use one process per "concern" per core (where applicable) and service as much as you can, as fast as you can, and just run like a freight train down through your priority queue.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
While I understand Intel is a CPU company, this seems to cloud their thinking. I'm not against an 80 core chip, but if it's idle most what good is it? What kind of I/O architecture is going to feed that many cores? Or are they simply going to copy Sun's Niagra/UltraSPARC-T1?
:)
(Amusingly, the CAPTCHA word is "critics".
Faster processors are great, but when will we see massive improvements in data storage...
About the same time that sound cards, input devices, shiny stickers on the side, and other COMPLETELY unrelated things catch up.
If it's for-profit but free, you're not the customer -- you're the product (e.g., the Slashdot Beta's "audience").
http://www.pouet.net/prod.php?which=9461 Try it out.
My blog. Good stuff (when I remember to update it). Read it.
But since this is /. versus Intel, everything has to be a joke, right?
Look dude, it took 10 years and K8 for people to stop calling AMD's products "crappy Intel clones", despite K7 being a pretty decent chip. It's going to take at least a few years before people, especially Slashdot, get over NetBurst and Itanium.
Some stuff allready does split into parallel jobs very well. Video processing (everything comes in discrete frames), multitrack audio processing, seismic data manipulation (really audio processing with a lot of tracks), finite element analysis - then there's just the option of running a pile of single threaded independant jobs at once. For some people it will solve the problem of getting all that scratch data to other nodes in a small cluster - you can just leave it in memory instead. If you have things that are dependant on stuff another CPU is working on then it gets hard because you are caring about how to make the solution as parallel as possible.
Smart voice recognition? Anyone tried it lately? Anyone tried to write pretty standard letters with it? Desastrous.
lol
AWESOME! I've always wanted a central heating unit capable of doing a gigaflop
To fix memory issues altogether, you need to move main memory onto the CPU wafer. This is a variant of the theme pursued by the processor-in-memory hardware advocates. If you alternate cores and memory cells and then use NUMA-like coding to make the memory look like a single unit (ie: the SGI Origin model), the number of cores directly connecting onto a piece of RAM would be very small - four cores to be precise - which is very manageable. Or, at least, more so than all eighty cores. You can also have simultaneous updates over many cells, which is not possible on a streamed bus. Ideally, you'd also use very high-speed transistors. This would be expensive, sure, but when you're talking 80 cores, you're already talking expensive. This ain't gameboy territory. In turn, this means L3 cache becomes meaningless and L2 would only be useful for remote memory accesses. My guess is that you could punch bandwidth up by an order of magnitude or better, by using parcelled-out high-speed RAM in this way. Well, SGI seemed to think it was worth it, as their entire brick technology was based on the concept, albeit at the macro level and not the wafer level.
Combining a switched network and local-wafer RAM, the machine would generate some serious performance. Serious heat, too - you'd need the combined wisdom of everyone on ExtremeCooling to be able to build a system that was stable. On the other hand, nothing on the macro-scale would come close this side of 2100.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
People focusing on parallelization will point to problems that are "embarassingly" parallelizable.
Let's assume that we have good memory behavior, or an interconnect like hypertransport or the as-yet-unseen CSI. Let's then look at something that can be broken into source and destination parts.
We'll go with image convolution.
So, we can break an HD image in to horizontal strips and work on them on separate cores, but we enjoy the cache locality of working on data within our range if we don't make our strips too small. Each thread working on an individual line in the image would be silly, as the overhead for the thread management would go up but cache coherency would go down.
Now, I brought this one up because I've actually worked on threading this very problem with OpenMP. Where does the knee of overthreading cost get really ugly on a four-core box when doing a 9x9 convolution? About 20 lines of HD material on my Woodcrest. That's 54 separate threads for HD material, if I feel so inclined. There's no reason to overthread so heavily, but it's doable without doing serious harm. If I had 54 cores actually doing the work, I'd have my image processing done in humorously short order.
Who would want that done by so many cores? Well, convolution isn't the only thing we do in our video editor, and I don't expect that it's the only thing that we will want to do in the future. I didn't even discuss task-level parallelism, but we have things like decoding multiple streams, pre-cooking expensive transitions, threading compression, etc. These things add up.
Other areas that come to mind just free-form typing?
Speech recognition, data mining, global illumination, traditional shading, ray-tracing, weather prediction, matrix inversion...
Let's look at matrix inversion:
Material science, physics, economics...
It's silly to go on. Being able to thread is hugely valuable to a large class of tasks. It opens up a lot of possibilities. Yes, it changes the way that you have to think about programming, but it's not wildly difficult to make the leap. I don't expect that everyone will be able to adjust (I've known plenty of decent programmers who couldn't reason through threading), but it's definitely coming. Hopefully we won't be wondering "aren't 200 cores 120 too many?"
However, that's beside the point. The "80" cores the article speaks of aren't general purpose cores where the operating system's scheduler could run just about any thread. They only do floating point calculations, which means that you have to write a certain amount of support code and your own scheduling and divide & conquer data I/O under any operating system to make any use of the cores. Linux is no more ready for them than Windows is.
Seriously, clusters suffer from a big problem - the I/O bandwidth is extremely limited compared to the CPU power. This means that sharing data over nodes (especially if you're using coherent memory) is where the problems lie. By shrink-wrapping the cluster onto a wafer, you eliminate the network I/O bottleneck, but introduce a whole bunch of other bottlenecks instead. (Ever seen 80 CPUs try to access the same piece of memory at the same time? The locking mechanism is a nightmare, the scheduling gives hardware engineers brain tumors, and avoiding being prosecuted for violating Amdahl's Law is enough to give any manager the night terrors.)
Really, at 80 cores, I don't believe Intel is capable of the imaginitive leaps required to beat the engineering problems. The problems are certainly solvable, but they are not conventionally solvable. Intel thinks conventionally far too often and their imagination tends to be filled with Cthulhu Mythos.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Gordon Moore must be proud of himself.
They're using their grammar skills there.
Eighty will only be useful to those rare people who need to run massively parallel algorithms.
640K ought to be enough for anybody.
I don't see anywhere in the article (other than the attention getting headline), nor do I see in the Intel intranet article, that there is a 5 year timeline for rolling out 80 cores. Where did this timeline come from?
I'll remember that next time someone is bashing GIMP or blender.
Hmmm. A hammer is for banging in nails and some people are better at it than others but that doesn't make the hammer an exculsive tool. You give an artist MSPaint and they'll produce better results than an amateur with photoshop, even if the amateur earns his 'professional' living using that software. Photoshop 'professionals' are 10 a penny and few are worth that, the tool is irrelevant.
Agreed that we need all the improvements you mention. But these smart technologies (voice or handwriting recognition, machine learning, et cetera) are mostly computationally-intensive, number crunching statistics tasks. The AI software you desire could benefit _greatly_ from multicore FP architectures.
......But Linux has been scheduling on systems with up to 1,024 processors already :)........
Scheduling is only a small subset of the larger problem. How do you break down tasks into parts that can be done simultaneously (parallel) and tasks that MUST be done sequentially and then put all that together into a coherent, useful result? Most thing people do and need done are sequential. Since computers are tools for people they must be taught to do things in parallel which are usually done serially. That is not always possible. The hardware is the easy part, It's the software that is hard. MS with their VISTA knows that by now.
All theory is gray
This is the console wars all over again.
Graphics don't matter! Shut up, you Wii fag, NO U, XBOX360 LOL, JAGGIES GIANT ENEMY CRAB RIIIIDGE RACERRRR, and so on.
Only this time it's Intel vs. AMD. Again.
Although I agree with you, I don't notice any performance changes apart from 3d videogames, the possibility of opening a divx file, or back when winamp used 30% of the cpu's resources just for playing a then-elusive MP3 file. We need smarter software and AI.
That fucking microsoft office paper clip doesn't count as an effort toward this, btw.
I, for one, welcome our new 80 core overlords
Great - can I have that in my Mac Mini with 2 TB of RAM and a 400 TB Hard Drive?
According to the article: "The chips are capable of exchanging data at a terabyte a second"
Obviously that information is not much use without more details though. I'd be particularly curious about the topography.
... like a metal disc sander for your face?
Maybe if you've got two CPU-bound processes, and you don't wind up running into bus bandwidth problems or memory contention issues, maybe you'll get almost the same performance, but I doubt it.
If you have the choice between switching between two threads doing data-stream processing (video processing, for instance) on a fast single-core or two half-speed dual-core procs, you can actually see better performance under some circumstances on the half-speed dual-core procs. Why? The scheduler is a general-purpose unit, not really written to optimally let one frame get processed entirely after the other. It has to work this way, because the illusion of responsiveness depends on some threads getting back to waiting in a timely manner, and the scheduler doesn't know which thread is which. This means that the two threads will switch back and forth on the single-proc machine, and you'll get cache eviction. This is far more important with modern prefetchers than the FSB or arithmetic performance (at least, for most of the image processing I do).
Obviously, the best solution is to have only one thread working at a time, coordinated to not contend for the resources of the processor, but we could be looking at two threads from separate processes. Coordinating these would be hard. This isn't just a hypothetical problem. I've done the legwork on this, writing the code, running profilers, timing execution, and working with optimization engineers processor vendors.
The rest of what you say is largely alright, but I'd encourage you to take a look at OpenMP (fork/join parallelism). Also, just building with nmake, switching from a 2p machine to a 4p machine (2.4GHz to 2.2GHz, even) cut my build time from 11 minutes to 7, and my computer is still usable for other tasks (like typing this message).
Would I want my MCE box to be 4p? Probably not, unless it meant that it ran at a lower clock/voltage and could pull off the same tasks while being passively cooled.
A modern graphics chip (GPU) has 48 fragment processors and a dozen more vertex processors - each of those does full scale floating point in four way parallelism at fairly respectable clock rates. So if all you need is a couple of hundred floating point units - you already have it.
The problem is the memory bandwidth into and out of this beast. The number of pins you can have is fundamentally limited - the bandwidth down each pin is fairly sharply limited. RAM chips are creeping up in speed V-E-R-Y S-L-O-W-L-Y compared to CPU speeds. So in the end, I think it's pretty much irrelevent how many processors you have on a chip because even with infinite compute power, if you can't get the data in and out, you're limited to RAM speeds.
The GPU's only get their speed and the ability to use all of that floating point horsepower 100% of the time (with no nasty Ahmdahl's law consequences) because they are working on Red/Green/Blue/Alpha or X/Y/Z/W in parallel - and millions of pixels are being computed using the exact same algorithm in perfect lockstep with few (if any) branches or loops. That means that a single instruction stream can feed all of those processors in parallel - and the data goes in and comes out in highly memory-friendly ways.
That's not much use for general computing though.
I glaze over when I hear about increasing numbers of processors - if the memory interfaces aren't designed very carefully PS-3 then very bad things can happen. Ask about how much RAM bandwidth this gizmo has.
www.sjbaker.org
The Connection Machine was a SIMD machine - Single Instruction, Multiple Datastream, which means that all the processors execute the same instruction at the same time, but on different data. Each processor had 4K of memory. There's a small class of problems for which this is useful, mostly big but localized problems like fluid dynamics and finite-element analysis. That class of problems was not big enough to support a market for the things.
Today's massively parallel machines are typically graphics processors. They have some resemblance to the Connection Machine, with all those shader units running the same tiny programs. But even they aren't SIMD machines; each little shader unit can potentially be running a different program. Giant SIMD lockstep engines seem to be unnecessarily inflexible today.
This reminded me, I was having a discussion with a coworker a little bit ago about how I wished that all of the popular compression algorithms would be implemented with multiple thread capability (the ideal would be either the application detecting the number of CPUs, or possibly an environmental tuning capability similar to "make -jN"). I did notice that winrar has implemented multiple thread usage in the more recent releases and even on my hyper threaded CPU which is not a true multicore system it has a noticeable increase in performance.
When will we start seeing all of the common UNIX compression utilities taking advantage of the great CPU scalability that many UNIX implementations and Linux like to brag about?
proxy
So that would leave us with Solaris.... Scales to 140+ cores today So if I just could get a usable GUI on top of it, or at least an up to date CLI......
(Sorry...it had to be said...)
96 cores, 10 watts, Intel has some catching up to do: http://www.clearspeed.com/products/csx600/
0 1.htm
About the people wanting to accelerate arbitrary functions, AMD's HyperTransport has the lead there: http://www.xtremedatainc.com/Products.html, http://www.fpgajournal.com/news_2006/09/20060906_
Take the most fundamental arithmetic and logic blocks, mix DRAM in locally, with massively parallel interconnections capable of asynchronous data transfer and data-driven clocking (i.e. the clock ticks as soon as the result is ready), and a compiler designed for the parallel environment which is race-condition aware. Basically, your design won't even run unless you have properly described the data dependencies. Add in some engineering training that teaches pipelining and concurrency. Replace stochastic scheduling with determinism. FPGA. Here's an example which I already linked elsewhere in this discussion: http://www.xtremedatainc.com/Products.html
All 80 cores will still be pegged rendering flash/java in a browser?
More things change, et al.
Have you read the moderator guidelines? Well, have you, PUNK? (and I want a Karma: Gnarly option)
Meanwhile, the only details that they ahve disclosed is that it is the size of an Eggo waffle.
.....and most importantly.....
Questions:
Power Consumption?
Heat?
Size?
COST?!?!?!
Efficiency to justify and astronomical cost?
Knowing Google's lust for data collection, the Soviet Union is still alive and well inside the psyche of Sergey Brin....
Most swe's don't know how to do parallel development, hence the crowd divides more.
Divide by Zero Exception.
Let's see... We have the following:
1 TB/sec total internal bandwidth.
1 trillion floating point operations per second.
1 floating point number = 4 bytes (at least; it could be 8 bytes, but probably isn't, since that would likely lower the marketing performance number).
How exactly can you do 1TFLOP of useful work with only 1TBPS of bandwidth?
Terra-operations - and my ISP upstream is still 128Kbps...
Yawnn
Don't forget Tcp Offload Engines, TOE for 10 GigE...
those will probably be pointless as well. It would be cool
to see how the interrupts are distributed to various
cores at a rate of 1 G interrupts / second...
Software radio becomes programmable without specialised
firmware... just raw transmission & recieve hardware connected to the mesh of cores. How will the FCC deal with that?
tv reception, cctv transmission & wlan on the same hardware...
cool!
They are claiming a terabyte per second interconnect. I think it is safe to assume it will be something like an Infiniband, myrinet or similar (NEC's IXS, IBM's HPS) high performance application networking technology.
What you're asking for is pretty standard stuff in the high end, where hundreds of processors is quite common. Cache coherency is a killer, and so they have died out long ago in the high end. when you think about it, CC basically requires a crossbar switch style memory archictecture which expands with the square of the number of processors, and much higher speed logic to resolve conflicts. So eventually, it doesn't scale. Instead multiple applications with large numbers of processors tend to only have small groupings (say 8 or so, but can go upto 30 odd) using shared memory/cc access) and then MPI for anything bigger.
Clusters have been using MPI for years for this sort of
thing. all the custom interconnects for supercomputing
have customized implementations in their MPI libraries to
take advantage of 1-sided communications. Most use a facility which can loosely be termed RDMA - remote direct memory access, another word sometimes used is OS-bypass. The idea is that for this sort of communications, you want to skip the TCP/IP stack and other OS buffering overhead, and just have straight memory to memory copies going on (under userland library control.)
folks generally don't do the direct invoking of things on other processors, but instead fire off jobs on blocks of processors, and have them communicate with 1-sided primitives. This is the sort of thing done on hundreds or thousands of processors today. It will just gradually percolate down to normal applications.
Hey great! Lets bring back Windows 95 for it!
I use to love the techno wannabees bragging about their dual CPU machines with win95 on them! Ah for the good old days!
How about a spell checker constantly running on any text you type? (OS X actually has this available but not all apps can/do use it) Or maybe you are tired of reading other people's misspellings and would like a browser that automatically corrects them. Sure, my computer could handle all of that now, but I'd rather it didn't have to choose between doing that and the multi-MB/sec transfers going on in Azureus (don't worry it's all legit)
Yeah, you may be right, but what he clearly ment to say is that it still takes the same amount of time to read slashdot every day during working hours!
molmod.com - computing tips from a molecular modeling
Mmmmm. Silicon wafer.
I think now the time has come to put the complete PC into one chip. With AMD buying ATI this could be a move into taht direction, since they could combine the GPU and the CPU into one chip. :-)
I predict that in future it will be possible to have the CPU, GPU, RAM and the harddisk (replaced by something like "RAM harddisk") into just one chip. That will be fantastic! No need for a huge gray box under your desk. Just plug the chip into your monitor and go for it
The condensated core will also be subject to the laws of quantum mechanics in that, before a program has finished running, there will be no way to know if it will crash or not. Microsoft plans to leverage this to further stablise their latest version of Windows. Security experts worried about the onboard "Quantum-Threading" technology redirecting portions of thread output randomly to other threads, were dismissed as not being "forward looking".
Meanwhile, AMDs new 1W, 128 core, 4098bit chip with 1GB L2 cache retails for almost 50% higher than Intel's Bose-Wintel chips, and has seen sluggish sales since the arrival of the new technology, despite its lower running cost that the 5MW Intel chip. When asked for comment, AMD's spokesman added; "Ch@#&t!! What the f**k is wrong with you people!??! Our chips save you money!! F@#*&^g cheapskates!!!"
Upon hearing the news, Linux founder and lead developer Linus Torvalds(51) said: "We're not rewriting the kernel for that monstrosity." Intel representative declared that the company was "dissapointed" in Torvald's remarks. Apple cofounder Steve Jobs(65), when asked whether Apple intended to release a the new Mac based on the chipset, declined to comment as he went about his daily 5km morning run. Apple pundits widely believe that the new Mac will run on a quad core Bose-Wintel Condensate, and to complement this will sport a blazing white, ultra smooth case made out of Bose-Einstien condensate, the fifth phase of matter.
In a related story, Microsoft cofounder Bill Gates(65), assaulted a technology reporter at a company press conference disccusing the new chip. Details are sketchy, but reports mention that one of Mr Gates older quotes about appropriate amounts of computer memory was brought up. Remond police have declined to comment on the case.
May the Maths Be with you!
"- We were promised Virtual Reality with VR Helmets more than 10 years ago - is this _just_ a matter of hardware?"
yes, the vr helmets are too expensive and uncomfortable. I agree with your other points but with this it's just the hardware (gamers would jump on this if it was not that expensive)
pkg-get install bash gnu-tools xfce4
enough said.
one...
two...
three...
presto!
(that trick never works)
Erlang already support multiple cores, it used to be that you had to start an erlang node for every core/cpu. Today a single erlang os-process will scale to the cores available.
s .html
/ part_frame.html
But from the programming point of view, erlang has supported multiple cores for as long as it has existed.
In erlang when you send a message to another process, you don't know if that other process is executing within the same OS-process or is executing in another OS-process, or even running on an distant machine. That is very good, because you don't have to rewrite anything to support distribution.
Concurrency is one of many features where erlang makes things easier for you.
On the other hand.. if you look at java or perl, those green processes or real threads (pthread) is just slapped on, like any other library. The language doesn't have any support for processes/threads nor is it oriented around processes/threads. This doesn't exclude you from doing erlang like stuff. It is just that an 10 line erlang-hello will expand to an 100 line pthread beast with sharp mutexes slashing around.
References:
New smp feature in erlang:
http://www.erlang.org/doc/doc-5.5.1/doc/highlight
Look at chapter 3, you don't have to understand erlang to follow:
http://erlang.org/doc/doc-5.5/doc/getting_started
The hardware to run Duke Nukem Forever
Karma: Bad. (As in Good?)
what about memory and other bottlnecks? Who gives a crap about having 80 parallel threads when they are all waiting for 1 resource (e.g. harddrive). Also, what about software? Writing software to take advantage of dual processor/core is tricky, and usually ends up being more buggy. Most software companies won't go back and redesign their software without being able to realize a profit margin increase.
to let my carts being drawn by 16 horses instead of 1024 chickens.
...
Somebody came up with this I forgot who.
I kinda wonder what sort of architecture they will use to make effective
use of this cluster without causing communication overhead slowing the
application down. I hope Intel has heard about Amdahl's law by
now, otherwise I'm guessing they are going to leave the field to IBM and AMD.
I just hate it like intel tries to optimize one thing only to peddle their
products to people who can only keep one thing in their mind.
Grrr
Je me souviens.
I'm concerned about the processing of my favourite porn clips, etc.
Is 80-core hardcore 40 times better than my current dual-core hardcore? That I gotta see!
Will this lead to more premature evacuations of my personal data?
Will we forever be saddled with the traditional 3-input architecture or will 80-core hardcore demand a new intercourse on how we interact with our orifice applications?
Finally, does Jennifer Love Hewitt care about how many cores I'm packing?
from: http://www.theonion.com/content/node/33930
Fuck Everything, We're Doing Five Blades
By James M. Kilts
CEO and President,
The Gillette Company
February 18, 2004 | Issue 4007
Would someone tell me how this happened? We were the fucking vanguard of shaving in this country. The Gillette Mach3 was the razor to own. Then the other guy came out with a three-blade razor. Were we scared? Hell, no. Because we hit back with a little thing called the Mach3Turbo. That's three blades and an aloe strip. For moisture. But you know what happened next? Shut up, I'm telling you what happened--the bastards went to four blades. Now we're standing around with our cocks in our hands, selling three blades and a strip. Moisture or no, suddenly we're the chumps. Well, fuck it. We're going to five blades.
Sure, we could go to four blades next, like the competition. That seems like the logical thing to do. After all, three worked out pretty well, and four is the next number after three. So let's play it safe. Let's make a thicker aloe strip and call it the Mach3SuperTurbo. Why innovate when we can follow? Oh, I know why: Because we're a business, that's why!
You think it's crazy? It is crazy. But I don't give a shit. From now on, we're the ones who have the edge in the multi-blade game. Are they the best a man can get? Fuck, no. Gillette is the best a man can get.
What part of this don't you understand? If two blades is good, and three blades is better, obviously five blades would make us the best fucking razor that ever existed. Comprende? We didn't claw our way to the top of the razor game by clinging to the two-blade industry standard. We got here by taking chances. Well, five blades is the biggest chance of all.
Here's the report from Engineering. Someone put it in the bathroom: I want to wipe my ass with it. They don't tell me what to invent--I tell them. And I'm telling them to stick two more blades in there. I don't care how. Make the blades so thin they're invisible. Put some on the handle. I don't care if they have to cram the fifth blade in perpendicular to the other four, just do it!
You're taking the "safety" part of "safety razor" too literally, grandma. Cut the strings and soar. Let's hit it. Let's roll. This is our chance to make razor history. Let's dream big. All you have to do is say that five blades can happen, and it will happen. If you aren't on board, then fuck you. And if you're on the board, then fuck you and your father. Hey, if I'm the only one who'll take risks, I'm sure as hell happy to hog all the glory when the five-blade razor becomes the shaving tool for the U.S. of "this is how we shave now" A.
People said we couldn't go to three. It'll cost a fortune to manufacture, they said. Well, we did it. Now some egghead in a lab is screaming "Five's crazy?" Well, perhaps he'd be more comfortable in the labs at Norelco, working on fucking electrics. Rotary blades, my white ass!
Maybe I'm wrong. Maybe we should just ride in Bic's wake and make pens. Ha! Not on your fucking life! The day I shadow a penny-ante outfit like Bic is the day I leave the razor game for good, and that won't happen until the day I die!
The market? Listen, we make the market. All we have to do is put her out there with a little jingle. It's as easy as, "Hey, shaving with anything less than five blades is like scraping your beard off with a dull hatchet." Or "You'll be so smooth, I could snort lines off of your chin." Try "Your neck is going to be so friggin' soft, someone's gonna walk up and tie a goddamn Cub Scout kerchief under it."
I know what you're thinking now: What'll people say? Mew mew mew. Oh, no, what will people say?! Grow the fuck up. When you're on top, people talk. That's the price you pay for being on top. Which Gillette is, always has been, and forever shall be, Amen, five blades, sweet Jesus in heaven.
Stop. I just had a stroke of genius. Are you ready? Open your mouth
This is the same crap Intel tried to sell us with their shared memory AGP video cards and soft modems back in the day.
... oh never mind, that's the Alpha and AMD's using Hypertransport already.
Accelerated hardware makes sense -- it may not be efficient in terms of transistors but its much more predictably efficient in terms of bottlenecks.
You have a CPU that sends off primitives to a GPU and you don't have a cache coherency problem in your calculations because the GPU is handling a completely different thing than the CPU and on its own bus with its own memory.
Now if you want to invent CPUs with multiple cores with multiple possibly independant switching pathways to main memory
- Michael T. Babcock (Yes, I blog)
In both cases, the code was building on a RAID 0 stripe of two drives. At least on my system, going to three drives made for a marginal improvement, and going to 4 did nothing (controller-bound for bandwidth).
We have a decently sized tree, and our build process builds orthogonal chunkcs of code in separately launched instances of nmake to improve performance. It's a simple approach to the problem, but it works well.
Oh, and the concern is blowing L2 cache for different image data on switches. We're talking about processing of two separate images (in memory). I'll try to dig up the code/numbers.
"So, we can break an HD image in to horizontal strips and work on them on separate cores, but we enjoy the cache locality of working on data within our range if we don't make our strips too small. Each thread working on an individual line in the image would be silly, as the overhead for the thread management would go up but cache coherency would go down."
Actually, you're dead-on talking about video encoding. A start-up company called Kula has developed a solution for using multi-core/multi-processor systems to speed up video encoding beyond levels that were thought possible. I saw their demonstration at IBC this year, and they had a 16-processor machine encode great quality H.264 SD using 2-pass VBR at over 6x realtime, which is ridiculous for a software solution. And their solution is scalable, so it could probably take advantage of 80 cores, given the proper attention.
Though, as I understand it, they don't break down the video spacially, but rather temporally (ie, each core is given a GOP to encode, then they are pieced back together). I'm not sure what this means for audio sync on longer movies, but these guys seem pretty bright. Hopefully they'll figure it out, if they haven't already.
Should be enough for anyone!
Archive is an even bigger issue. Unless, of course, you just get 2 hard disks and use one to back up the first. But we really need TB size optical at a reasonable cost to make a complete backup of a system and we should be able to do a complete bare metal recovery on the system from the backup media.
(Office of Craig Barrett, Intel CEO, circa 2010...)
Would someone tell me how this happened? We were the fucking vanguard of semiconductors in this country. The Core Duo was the processor to own. Then the other guy came out with a quad-core processor. Were we scared? Hell, no. Because we hit back with a little thing called the Core Octa. That's eight cores and an co-processor. For cryptography. But you know what happened next? Shut up, I'm telling you what happened--the bastards went to ten cores. Now we're standing around with our cocks in our hands, selling eight cores and a co-processor. Cryptography or no, suddenly we're the chumps. Well, fuck it. We're going to eighty cores.
Sure, we could go to sixteen cores next, like the competition. That seems like the logical thing to do. After all, eight worked out pretty well, and sixteen is the next logical step after eight. So let's play it safe. Let's make a faster memory controller and call it the Core DecaOct Extreme Edition Pro. Why innovate when we can follow? Oh, I know why: Because we're a business, that's why!
You think it's crazy? It is crazy. But I don't give a shit. From now on, we're the ones who have the edge in the multi-core game. Are they the best a geek can get? Fuck, no. Intel is the best a geek can get.
What part of this don't you understand? If four cores is good, and eight cores is better, obviously eighty cores would make us the best fucking processor that ever existed. Comprende? We didn't claw our way to the top of the processor game by clinging to the single-processor industry standard. We got here by taking chances. Well, eighty cores is the biggest chance of all.
Here's the report from Engineering. Someone put it in the bathroom: I want to wipe my ass with it. They don't tell me what to invent--I tell them. And I'm telling them to stick seventy-two more cores in there. I don't care how. Make the cores so small they're not functional. Put some on the leads. I don't care if they have to cram the last ten cores in on the other side of the die, just do it!
People said we couldn't go to four. It'll cost a fortune to manufacture, they said. Well, we did it. Now some egghead in a lab is screaming "Eighty's crazy?" Well, perhaps he'd be more comfortable in the labs at NVidia, working on fucking GPUs. Vertex shaders, my white ass!
I know what you're thinking now: What'll people say? Mew mew mew. Oh, no, what will people say?! Grow the fuck up. When you're on top, people talk. That's the price you pay for being on top. Which Intel is, always has been, and forever shall be, Amen, eighty cores, sweet Jesus in heaven.
[Apologies to the Onion.]
"Ladies and gentlemen, my killbot features Lotus Notes and a machine gun. It is the finest available."
It's official... I don't understand your post. :-) Thanks, though.
They've been pimping it the last 2-3 years at their booth at the Supercomputing Conference - note that I'm not sure whether it's been released or not, because other than staring at the awesome projection units they use and laughing at the MS employees trying to boot up their systems ( nothing makes you realize how bad the windows guys have things until you watch them trying to reconfigure one of their systems and arbitrarily failing, though 2K/XP is an order or magnitude better than their previous attempts at an operating system ), I stay the hell away from that booth.
PC moderators can suck my White pierced, tattooed dick. If you think pride == hate, s/dick/Aryan meat mallet/g.
80 cores? No problem. Imagine a programming language that calls EACH THREAD on a different core. Works for me. Hell, I'm running 34 processes right now. If I could get 34 cores to run them I'de be plenty happy. Hell as far as I'm concerned, once we get started 80 just wont be enough. Someday bill gates will say "Whos ever gonna need more than 640 cores?"
um... I guess it ain't clear then... the parent post was saying that you need OS support for accelerated remote procedure calls and one-sided communications. However, 1 sided communications already is in standard use by folks using hundreds of processors through an already
standardized library: MPI - Message Passing interface. Rather than the OS needing to define a new API, the folks creating high speed interconnects just create optimized libraries (in order to sell their hardware). Folks writing codes for hundreds of processors tend to want to treat them as array elements, so the chaotic calling of procedures just is not that useful. so the RPC support he is asking for is not really important.
In other words, the software stack for using large numbers of processors is already well-known. No need for any new OS features.
Noone has proved that all serial processes can be done in parallel (and most people think they can't), and that's independent of the language, so your post is nonsense unless you have a demonstration that complexity classes NC and P are equal (which would be great if it happened).
Haskell is not more powerful than any Turing complete language (as is demonstrated by the fact that it runs in turing complete machines (computers).
The AACS key is NOT 0xF606EEFD628B1CA427BEA93A9CA9773F
Thanks. I got the gist of it, just not all the details. P.S. Firefox 2's HTML forms, real-time spelling checking is great. I was going to spell it "jist".