Posted by
Hemos
on from the conjecture-or-convinced dept.
vmircea writes "If you think clock speed is the most important measure of a processor, IBM's Bernie Meyerson wants you to reconsider. Meyerson, who heads research and development efforts for Big Blue's semiconductor group, says processor chip speed is old news. Go to ZDNet for the interview."
296 comments
Speed
by
Anonymous Coward
·
· Score: 5, Insightful
The end result that people care about. When a system is purchased,
and people are looking at transaction processing capabilities, that is an end
result. They are not looking at whether the clock frequency of the
microprocessor is 8 percent higher.
Isn't that how non-idiots have been looking at it, all along?
I don't think this is really a new attitude.
Re:Speed
by
Anonymous Coward
·
· Score: 4, Insightful
agreed.
Except taht most people stop their research into what chip they want when they see a RBFN with the letters "MHz" or "GHz" printed next to it. Nevermind how other factors influnence true optimality of a chip. I personally would much rather see a standard numerical rating be developed (FLOPS may work), except that some (coughIntelCough) wont use taht in marketing materials because it shows inefficiency. (Much like how Hummers dont print their gas mileage on showroom display materials)
FLOPS won't work; it ignores workloads that use integer math. It also ignores workloads that specialize in vector math. And workloads that depend a great deal on automated decision-making. And random-number generation.
The problem is that no matter what metric you use, it won't fit all cases. Different workloads have different requirements. Personally, I'd like to see programmable hardware...Essentially an FPGA section on CPUs. Programs would provide the OS's scheduler with a circuit layout, and the scheduler would have the layout programmed in when needed.
Each program doesn't necessarily have to have access to the whole grid array, either. The scheduler could divide the array into sections. One section would be for speeding up scheduler operations. The rest would be available to have programs loaded in. You wouldn't even need to erase one program's hardware when another program had something it wanted to implement. With the hardware divided, you could load the new program's code into an empty slot, and leave the old code available for the old program's next timeslice. (To prevent having to reprogram the FPGA section every time the program's turn came about.)
SPECfpBase2000 and SPECintBase2000 cover almost everything usefull you can do with a computer. Btw the problem with FPGA's is that gate reconstruction times are so slow that you REALLY have to be doing a lot of something for it to make sense, like compressing an entire movie, and even then you might be able to do 90+% of the speed by using assembly and vector ops. Add to that the fact that FPGA's have a limited number of rewrite cycles and generally uses different fab processes and it gets to be pretty stupid for most applications.
-- There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Re:Speed
by
Anonymous Coward
·
· Score: 2, Interesting
Oddly enough, this idea sounds like a research paper I read... like 4 years ago, now; and I'm sure there were more earlier, as the paper referened other similar adaptable architectures.
The drawbacks: you either need a FPGA that you can reprogram very quickly (on the order of nanoseconds) or you need a task that takes a long time and can more than benefit from having to take the time it takes to reprogram the FPGA. The latter is not terribly useful for your average desktop machine.
Most FPGAs are RAM-based. Reconfigure as much as you want. This includes every Xilinx FPGA made. And there are some Xilinx Spartan II parts under $10. Pretty cool!
There are only a few FPGAs which use any sort of non-volatile memory (Actel's Pro-Asic being one). Those would have a limited life.
-- "-1 Troll" is the apparently the same as "-1 I disagree with you."
The drawbacks: you either need a FPGA that you can reprogram very quickly (on the order of nanoseconds) or you need a task that takes a long time and can more than benefit from having to take the time it takes to reprogram the FPGA,/i>
That's why you'd divide the gate array into sections. Assuming you weren't trying to run too many apps with FPGA code at the same time, you can leave the section of code in there while waiting for that process's next timeslice.
"Most people" aren't interesting. "Most people" won't be missed. And I don't really care how satisfied "most people" are with whatever computer they end up buying.
Except taht most people stop their research into what chip they want when they see a RBFN with the letters "MHz" or "GHz" printed next to it. Nevermind how other factors influnence true optimality of a chip.
I think you're misreading the article. It isn't (mainly) decrying the "MHz Myth."
What it is saying is that the expectation of exponential growth in processor speed (not just MHz) can no longer be maintained. Processors are not going to improve as fast as in the past. So they are trying to turn attention from the CPU, and optimizing the computer at other levels, up to the level of "how much will it cost to obtain and support a payroll system?"
meh this is why I'm not allowed in the geek circles at school, I tell these people that cpu speed in MHZ doesnt count, and it's how it performs the tasks and how efficient it does it that counts, they call me a dumbass because "cpu speed makes it more efficient, dumbass!"
also, they think windows is the most superior OS ever, and bill gates is the inventor of the modern computer, etc and that he has the highest IQ in the world.
Personally, I'd like to see programmable hardware...Essentially an FPGA section on CPUs. Programs would provide the OS's scheduler with a circuit layout, and the scheduler would have the layout programmed in when needed.
sure... great idea. just like the idea of putting an array of dsps on the pci bus, to act as a generic accelerator unit.
of course, the problem with both of these is: who is going to program for these? some specialized high performance applications, perhaps (openssl, linux kernel, autocad,...). will "grep" or "apache" make use of these accelerators? can they be programmed in c? are they portable? do you have to be a "ibm power fpga programmer" to know how to use these, or can any programmer easily port their skills here? having programmed dual core risc+dsp chips, i will tell you, the complexity of splitting a task between the risc and the dsp is much much much much greater than doing in on the risc alone. when you distance yourself from general-purpose computing, you lose the support of just about everyone.
These people are not geeks my friend. Sadly, you seem to have encountered the rare, but highly dangerous wannabeus geekus. I would advise you to vacate the area immediately, as they frequently attack without any warning whatsoever, spraying their deadly stupidius maximus poison directly into one ear, where is promptly destroys all brain tissue it comes into contact with, and flows out the other ear. Run. Run fast. Run hard. and whatever you do, DON'T LOOK BACK.
FLOPS won't work; it ignores workloads that use integer math. It also ignores workloads that specialize in vector math. And workloads that depend a great deal on automated decision-making. And random-number generation.
That's all fair but it's a sight better than MHz/GHz - at least it's a measure of the CPUs ability to do useful work of some kind. The current numbers are about as useful as RPM as a measure of motor vehicle performance - relevent but not the whole story.
I'm pretty sure Heinlein said it through Lazarus Long, but Heinlein also has a habit of uncredited quotes. Or even partial-references ("Not mine...some ancient Earth bloke.")
I figured the software compiler would handle it. Add some extensions to translate certain common bits of code into FPGA code.
Shared libraries would especially benefit. FFTs and regex engines, for one. SIMD trigonometry functions. A Perl or Java VM. The 3D accelerator's software driver. CRC checking.
Some kernel modules would benefit as well. Framebuffer code, crypto ciphers and digest algorithms. The scheduler (as I mentioned earlier). The IRQ handler.
Even if code doesn't necessarily require superfast algorithmic processing, you could treat FPGA space as another cache. The kernel could provide a secure IPC mechanism that kept all data on-die.
This is truly a case where if you build it, and provide some practical sample code, they will come.
Ask the damn question!
by
Anonymous Coward
·
· Score: 1, Interesting
"If you think clock speed is the most important measure of a processor, IBM's Bernie Meyerson wants you to reconsider."
At no point do I see him ask "If clock speed isn't important, what is?" Of course it gets answeredin a round-about kind of way.. but still.
Re:Ask the damn question!
by
Anonymous Coward
·
· Score: 0
Did he say it wasn't important? Don't put words into other people's mouth. He said it wasn't the most important measure of a processor. And he is right.
Re:Ask the damn question!
by
An+Onerous+Coward
·
· Score: 2, Interesting
I think the thing he was trying to focus on was their attempts to implement a "reconfigurable" or "programmable" chip. If this could be done well, I'm guessing it would take the whole "processor speed doesn't matter" thing to a whole new level, above what they teach in a basic computer architecture course.
Remember back when all the Pentium chips had "MMX" slapped onto the end of their names? MMX stood for "MultiMedia eXtensions," which were a set of operations programmers could use to speed up certain highly repetitive tasks like manipulating video or audio streams. I'm not familiar enough with them to provide a good example, but the point is that it is possible for a chip designer to implement a chip in such a way that it performs very well at certain tasks.
Make the chip configurable, and suddenly all this power falls into the hands of the programmer. So there would be certain situations where a 100MHz processor using the proper configuration could seriously outperform the same chip at ten times the clock speed, if the second was using a more general configuration.
--
You want the truthiness? You can't handle the truthiness!
Re:Ask the damn question!
by
qwerty101
·
· Score: 1
Why not do what was done in the old days - benchmark the system you want to buy?
Wow, what a paradigm shift
by
J.+T.+MacLeod
·
· Score: 1
They're only, what, almost a decade late making the observation that it's no longer as relevant for the average consumer?
Sorry, I'm a bit bitter today.
Re:Wow, what a paradigm shift
by
Sir+dies+alot
·
· Score: 3, Funny
Well, it is IBM. In the past, their ability to think 'ahead' was rather well portrayed when they decided that there was no home market for computing. But at least they are catching up right?
-- The stupidity of your average American is just about the same as the average European, we simply show it off better.
Re:Wow, what a paradigm shift
by
JazzHarper
·
· Score: 2, Insightful
I beg to differ. It's relevant to the self-esteem of the average consumer, when comparing his purchase to that of his neighbor.
I find it ironic that MIPS were dropped from most advertising, in part, because they were misleading, so manufacturers went back to quoting clock frequency, which is even moreso. --
Re:Wow, what a paradigm shift
by
the_2nd_coming
·
· Score: 1
actually, it is only about 3 years or so out of date.
--
I am the Alpha and the Omega-3
Re:Wow, what a paradigm shift
by
J.+T.+MacLeod
·
· Score: 1
The days of the 486DX4 versus the Pentium come to mind...
Yeah, but it's such a pain in the ass to find the right size CPU shim to hold a pot, and to top it off, you've got to find the right angle so the pot handle fits inside the case. Plus we all know how Intel cheated on the benchmarks by using Grade A jumbo eggs when measuring their competitor's products. Don't even get me started on the whole fine print thing with "hard-boiled" and "soft-boiled". Never trust a benchmark!
-- Fred
"A fool and his freedom are soon parted" -RMS
Re:real speed
by
Anonymous Coward
·
· Score: 0
Funny, I measure the speed on how many eggs I can cook around it per minute. Who says you have to put the eggs on the chip?
Look, you can't benchmark an Apple that way. What the heck are you thinking? You benchmark an Apple by baking a pie on the CPU. After it cools somewhat, and you've had your way with it (for er... benchmarking... purposes), you conspire with your old man to tell your mom you ate it, dig?
No, it's the speed holes that make it go faster. You know, those tiny little holes in the socket -- the more holes, the faster the processor.
And I'll let you in on a little secret -- those pins that go in the holes are actually there to slow the CPU down. No need to buy a new processor -- just clip off a couple of the pins on your current PC and it'll go much faster.
And I'll let you in on a little secret -- those pins that go in the holes are actually there to slow the CPU down. No need to buy a new processor -- just clip off a couple of the pins on your current PC and it'll go much faster.
I tried, but the magic smoke came out of the CPUand it doesn't work anymore (for those who knows, the magic smoke is what makes the chip work; if it escapes, the chip dies).
Right on about processor speed not determining how fast a computer is.
Why do people call the clock frequency for the "processor speed"? For me processor speed is how a processor performs in benchmarks. Calling the clock frequency speed is like calling a car's RPM for speed, when the obvious real measure is how fast the car is actually going.
No love for the Daewoo Nubira Type R? Hata! I wizzle yo shizzle
I agree with Bernie
by
L.+VeGas
·
· Score: 5, Funny
Clock speed is good, but what I look for in a processor is that ephemeral processor attitude. Can I show it off to friends? Will my mother thinkk it's cute, or is it a little... dangerous? I want a processor that says something about me. That I'm a rebel that won't take no for an answer. That I'm cool without trying. If a processor can't do that for me, well I'm just not interested.
Re:I agree with Bernie
by
foidulus
·
· Score: 4, Funny
Clock speed is good, but what I look for in a processor is that ephemeral processor attitude. Can I show it off to friends? Will my mother thinkk it's cute, or is it a little... dangerous? I want a processor that says something about me. That I'm a rebel that won't take no for an answer. That I'm cool without trying. If a processor can't do that for me, well I'm just not interested.
I have a suggestion, get a personlized tattoo on your ARM! *Rim shot
I apologize.
Re:I agree with Bernie
by
hey!
·
· Score: 2, Funny
I have a suggestion, get a personlized tattoo on your ARM!
Hmmm. What Bernie is desdribing here sounds like an Adaptive Semiconductor System.
-- Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Re:I agree with Bernie
by
Anonymous Coward
·
· Score: 0
Indeed, Sir! Tattooing your ARM or (StrongARM) sounds much cooler -- so much more *dedication* -- than just slapping a removable sticker on your case. Do you think the Z80 in my own PDA would look better with a personalized tattoo?
Poeple still want more ghz...
by
mw5299
·
· Score: 2, Interesting
Its interesting that apple is releasing (in july) the IBM made g5 that can go to 2.5ghz. It seems like people still care if a prosessor can go "more ghz". I think it is smart what AMD did with there 3000+ chips(or an other somthing+ chip). It makes poeple think that prosesor runs faster when it realy doesn't.
Re:Poeple still want more ghz...
by
Anonymous Coward
·
· Score: 0
Is it July already?
Nice timing.;)
Re:Poeple still want more ghz...
by
Short+Circuit
·
· Score: 3, Insightful
What bugs me is how AMD and Intel kicked up their processor speeds. Both of them made their pipelines deeper.
While that's fine for some workloads, with more instructions being executed at the same time, it harms workloads that depend heavily on the results of current calculations to figure out what to do next.
Intel eased the problem by implementing hyperthreading; I'm surprised we haven't seen the same thing come out of AMD's corner.
Re:Poeple still want more ghz...
by
Jameth
·
· Score: 5, Insightful
"It makes poeple think that prosesor runs faster when it realy doesn't."
Actually, it makes people think the processor runs faster when it really *does*. Which is why I like their numbering scheme: it compensates for consumer ignorance.
Re:Poeple still want more ghz...
by
AstroAndy
·
· Score: 0
Up until about the AMD 2200 XP+ the numbers actually meant something. I have a AMD 2000 XP+ on my computer which was almost the best availible at the time (2100). The numbers then meant that although the clock speed of the chip wasn't 2.0GHZ (in my case 1.4Ghz), it operated as well as a P4 @ 2.0GHz. As we have talked about before clockspeed doesn't mean that much if your architecture sucks. I always have felt that AMD's architecture was far superior to Pentium's. Of course after around the 2200, AMD decided to release processors with insane numbers just to screw with people's heads. Oh, and BTW, by processor still suits my needs very well.
Re:Poeple still want more ghz...
by
Anonymous Coward
·
· Score: 0
AMD's pipeline is not that long. Hyperthreading wouldn't do any good.
Re:Poeple still want more ghz...
by
Anonymous Coward
·
· Score: 2, Insightful
It makes poeple think that prosesor runs faster when it realy doesn't.
Partially. When I bought my 2200+, I understood that it ran at 1800 MHZ. I looked at all the specs first. And then I saw the comparison to the Intel machines that I was looking at. For less money, I bought my AMD parts that put the Intel equivelent to shame. At 1800 MHZ, the Intel couldn't come close. At 2200 MHZ, as AMD wants you to think of when comparing chips, they were pretty close. The Intel chip beat the AMD in a few catagories. But I wasn't prepared to drop the extra $200 at the time for the few improvements getting the Intel would provide. That was the why I went with AMD...better value and performance for your money.
Re:Poeple still want more ghz...
by
xplosiv
·
· Score: 2, Informative
Actually, the rating means how fast the AMD XP cpu is compared with the original Athlon series, not the speed of the P4 counterpart it can compete with. You can read more about the XP rating in the AMD press releases when they first introduced it, search their site or use google.
Re:Poeple still want more ghz...
by
ThisIsFred
·
· Score: 1
A 2.4GHz Opteron would be more than fast enough for me. What I'm looking forward to is the ability take advantage of the bandwidth available to, say, DDR PC4200 RAM. Seems like we've stalled at 200MHz (PC3200).
-- Fred
"A fool and his freedom are soon parted" -RMS
Re:Poeple still want more ghz...
by
dasmegabyte
·
· Score: 3, Funny
Then maybe you want to buy the new dasmegabyte XELEXONIUM 4500+ processor. It may look like a ram chip I pulled out of an old TI calculator, but trust me: with two Xs in the name and a plus after the 4500, it MUST be fast.
Also, those drinking straws glued to the top are longer pipelines.
Re:Poeple still want more ghz...
by
Anonymous Coward
·
· Score: 0
Screw the P4. The numbering scheme refers to an AMD Duron at 1000 MHz (i.e., the "1000" reference CPU).
AMD uses a largish suite of application benchmarks to measure the rating of each new CPU. It is by and large agreed on hardware forums that (even with their latest models) their rating is very close to the truth -- in other words, they have kept it reliable.
It's a nice coincidence (but no more than that) that a "3400+" races head to head with a P4 3.4GHz.
Re:Poeple still want more ghz...
by
CriX
·
· Score: 1
Why don't they just start rating their chips by peak, sustained MFlops?
That's a pretty universal scale right?
Maybe that would be too direct, and would undermine the expensive, shiny and flashy advertising.
--
Moderation: +1 pwnage
Re:Poeple still want more ghz...
by
cft_128
·
· Score: 1
Why don't they just start rating their chips by peak, sustained MFlops?
That's a pretty universal scale right?
Maybe that would be too direct, and would undermine the expensive, shiny and flashy advertising.
Only 'universal' if you do not care about integer operations or how the chip deals with branch prediction failures and pipeline stalls. Vector operations can also be handy...
The problem is that there is no universal measure. There are many benchmarks out there but they all target a specific area. The best way is to benchmark the chip based on what you are going to be doing -- many magazines that test computers will use Adobe Photoshop and MS Office as part of the tests, thus giving the reader a better idea on real-world performance.
This of course is not just a measure of the CPU but the whole computer - Graphics card, disk drives, memory, how the motherboard does IO, etc.
My clock speed is 64k hz
by
Anonymous Coward
·
· Score: 3, Funny
if you think clock speed is the most important measure
I think it is very important for clock speed - the crystal in my watch runs at 64k hz to keep time which is quite important. Lets see you Solaris or AMD overclockers beat that!
Re:My clock speed is 64k hz
by
jrrl
·
· Score: 2, Funny
I've overclocked my watch to 67.2 khz and it rocks! My workouts go much faster now (and better what with the water cooling rig acting as a wrist weight).
Although I never seem to get enough sleep and nobody ever shows up on time for my meetings... Hrm...
-jrrl.
>I've overclocked my watch to 67.2 khz and it rocks! My workouts go much faster now (and better what with the water cooling rig acting as a wrist weight). Although I never seem to get enough sleep and nobody ever shows up on time for my meetings
Let's see. Less exercise, less sleep, less meetings.... that means more games. I fail to see the problem here.
-- Tired of legitimate data sources? Try UNCYCLOPEDIA
What else besides games?
by
MacFury
·
· Score: 4, Interesting
What other applications besides games really tax the CPU right now?
I do a fair amount of video editing and image manipulation, even still my two year old computer works fast enough for me...
Does the average Joe need the computing power they are given?
I dabble in image manipulation, but mainly i do music production on my computer. Mine's about 2 years old, too...Most of the time, it does great. However, once i get about 7-8 tracks in, depending on the kind of output, plus effects like reverb and compression, my CPU says, "Aw, HELL no!" and freezes my machine. Happened just yesterday..makes me wonder sometimes if 2 proc's might help.
Re:What else besides games?
by
southpolesammy
·
· Score: 1
Having 2 or more CPU's will only help you if your application has the ability to use them. If it's a single-threaded app, then you'll just be heating up your system case more than necessary for no gain.
Not exactly. Using SMP helps in other ways. The OS/resources need managing and when his application is using CPU 1 at 95-100% CPU 0 is hugging curves and taking names at about 30% doing memory and disk management, etc. He is having lock-ups because he has one cpu. More RAM, faster disks, more CPUs will help him out. While the application may not be able to make use of multiple threads (hypothetically, it very well might be able to) the system as a whole sure could use multiple thread execution. HT is not the answer as they both rely on the same cache and also similar parts of the same CPU.
> However, once i get about 7-8 tracks in, depending on the kind of output, plus effects like reverb and compression, my CPU says, "Aw, HELL no!" and freezes my machine.
Heat problem. You CPU freezes because it is too hot.
Re:What else besides games?
by
iwadasn
·
· Score: 1
I'm currently reorganiznig about 300 gigs of data (zipping and unzipping), I think that counts.
Re:What else besides games?
by
demachina
·
· Score: 1
Paint. Some effects can be fairly intensive.
CAD and architecture.
Animation and modeling, the interactive 3D part and rendering.
Video editing.
Flash or animated SVG.
Virtual reality, either games or some kind of virtual presence. There is never going to be enough CPU or graphics power to accurately model reality or unreality. As the CPU, GPU and memory continue to grow so will the level of realism and complexity of the virtual world. Snow Crash and Diamond age laid down the gauntlet, its been slow in coming, but its inevitable people will start moving part of their life in to virtual worlds where they can assume alternate personas and interact with people around the world with out the constraints of physical space or the crappy meat bag they are stuck with.
Speech recognition.
AI.
HDTV DVD, OK so todays CPU's can handle this, then double the resolution.
A host of science and engineering applications like CFD, seismic, finite element analysis, PCB layout, chip simulations etc. though its a little way out where people might design and manufacture custom widgets in their homes.
Dude, the POWER chips are processors which are designed for SERVERS. It's a different market. You could say in the server market, I/O performance trumps processor power (I'm not only talking about networking and disk I/O but also I/O from main memory to the CPU). However, in many applications, CPU power is necessary. For things like scientific calculations, or even business applications like data mining.
Not saying that everyone needs it, but there is definately a market for it.
Moreover (I should add this) every once in a while a new application comes along that hogs CPU power. One example, just 10 years ago you needed all the processing power you could get to play MP3's. MP3 players were hand-coded in assembly, and you need to go down to DOS to run them. Now playing an MP3 is a background task. Any one can play MP3's, and based on the popularity of Napster a few years back, indeed everyone was playing MP3's. The bottleneck was network bandwidth, no one even thought of processing power as a problem.... but this would have been the problem just a few years earlier!
With processing power, if you build it, they will come. Here's the kicker: they won't even notice that they're using the extra processing power. The applications just weren't that common before...
Re:What else besides games?
by
Junks+Jerzey
·
· Score: 1
as a developer i want compilation to be a quick as possible.
Then don't use C++. Seriously.
I've been using Delphi since it was Turbo Pascal, and compile times have been a non-issue for me since I can remember. Even on my 333MHz P2, I've never had a perceptible compile/link time, ever. Even for full rebuilds.
And these days, anyone doing heavy software development should be using Perl, Python, Lua, Lisp, anything dynamic and lightweight.
Re:What else besides games?
by
ThisIsFred
·
· Score: 1
as a developer i want compilation to be a quick as possible.
If you're doing x86 stuff, you want an
AMD processor.
Most likely running out of RAM. Once the system starts paging to disk, you're screwed.
Re:What else besides games?
by
DrAegoon
·
· Score: 1
Obligatory MS Bashing...
If Microsoft would take a release to spend on fixing bugs and performace issues instead of adding more "features" to convince Joe user he needs a new machine, people would start getting better use out of their processors. Instead MS decides to add these "features" on top of its already busted code sapping performance and making their existing problems harder to fix.
Want to see for yourself? Try running Windows 98 and 98 era software on a current machine, if you can get past the constant crashes, the performance will seem faster than current software on XP on the same machine.
Most audio production apps can take advantage of SMP. Even if his can't the extra CPU will be used by the OS for I/O and system processing. Add to that the fact that the most common CPU hogging thing is effects modules which are seperate processes and you can see where SMP helps. I just built a dual Opteron rig for a friend of mine, only 4GB of ram but he has slots for 4 more GB and he already has beta 64bit plugins. The machine I built him should last him until the hardware starts to get flaky in 3-5 years.
-- There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Re:What else besides games?
by
Egekrusher2K
·
· Score: 1
Video encoding. Real time manipulation of video data is still impossible with even the fastest of desktop machines. "image manipulation"- I take it you're talking about Photoshop. That depends almost entirely on the amount of physical RAM in your system. It is not very processor dependant at all.
There are also things like rendering animation (which can take weeks on even high end PCs), large math calculations, voice recognition, etc.
There are LOTS of things that would take advantage of faster processors.
-- Listen to my experimental-industrial-techno!
Re:What else besides games?
by
Bert64
·
· Score: 1
The sourcecode to quake 3 was released? when? where do i get it?
Re:What else besides games?
by
Bert64
·
· Score: 1
Dynamic perhaps, lightweight no.. scripting languages have far more overhead than compiled code, you have to run the interpreter aswell as the actual program, you consume more ram and more cpu every time you run the program, as opposed to just once when you compile it.
Scientific computing. Here are things I am working on now that I wish I had more power for:
Artificial vision - I wrote and use analysis software that tracks motion of magnetic poles on the Sun's surface. There are about 10,000 - 50,000 poles visible on the surface at any given time. It takes a full day for my dual Athlon to process a full day's worth of data. A 10x speedup would be great!
Space physics simulations - trying to describe the behavior of solar storms is very computationally intensive. It takes a minimum of ~1016 floating-point operations to simulate even a simple coronal mass ejection.
That's a CPU-year.
Sound reprocessing - as I digitize my LP collection, I depop and noise-gate every track. It takes 2-3x as much clock time to digitally process the music as it does to play the LP into my sound card to be digitized.
Compilation - enough said
Stupidity in scientific software. A lot of my scientific work involves one-off codes to perform a particular operation on my data. Robust, reliable, transparent algorithms are just as important as the final result (after all, how can you trust a result if you haven't spent the time to understand the algorithm and all its nuances)? The stupider the algorithm, the better. If I can save a day of analytical work with a CPU-hogging numerical solution that takes 10 minutes to code and an hour to run, I've saved most of a day -- unless I have to run that code more than ten times. But as my CPU gets better the tradeoff improves and I can get my work done faster. The savings here isn't in direct CPU time, it's in allowing me to use stoopider code more of the time.
I'd really like an AI to write/. replies for me so I can get more work done.
What other applications besides games really tax the CPU right now?
Keeping a Gentoo system up-to-date?;-)
GNU Radio is a great example of something that is very cool, but requires more processing power than most people have. It's right on the horizon of feasibility (so it's not just a pipe dream totally out of reach) but still makes you ache for more power, more, more, more!!
-- As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
Re:What else besides games?
by
Dan+Ost
·
· Score: 1
Image Processing.
For example, in order to submit images for revenue processing, we have to convert our beautiful grey-scale images into binary images. In order to get the least ugly result possible, we use some pretty sophisticated and cpu intensive algorithms. On a 2.4GHz Intel processor, conversion takes about 10 seconds per image which means it takes about 8 cpu-hours to convert a typical night's worth of images (4 actual hours thanks to our dual-proc machines). On high volume nights, the extra time can push us past our submission deadline. If we could shave 1 second off of each image, that would move us out of the dangerzone.
That said, for most of what we do, I/O is the real bottleneck.
--
*sigh* back to work...
Re:What else besides games?
by
the+morgawr
·
· Score: 1
A long time ago. The game source was released so that modders could use it to make mods. What they are compilling doesn't have the engine (it's part of a separate library).
-- The policy of the United States is worse than bad---it is insane. -- Ludwig von Mises, Economic Policy(1959)
Re:What else besides games?
by
drinkypoo
·
· Score: 1
vmware depends much more on memory bandwidth and memory usage than CPU utilization unless you're peforming CPU-hungry tasks in those virtual machines. The second best thing you could do for your vmwares is get an Opteron with a dual 400MHz DDR memory controller. The first best thing would be to get several of them:) Since all the peripherals are virtual the majority of memory bandwidth needs to be to the CPU and not the bus anyway, which is ideally suited to a system with the memory controller integrated into the CPU.
-- "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Re:What else besides games?
by
drinkypoo
·
· Score: 1
Amusing you should mention both games and CFD but not in the same paragraph. Imagine what it will be like when we can use CFD to model fluids in games, the water effects will be astounding. Also, the more CPU you have, the more generic you can make your physics models, until one day you find yourself able to use the same physics for boats, planes, aircraft, and so on. It would be fantastic to be able to model the behavior of each bushing, shock, fuel injector, et cetera. I want reality! Of course, I can step away from the computer and experience real reality (redundant, I know) but in the really real world when you crash into a tree at 160 MPH bad things happen, both to you and the tree (but mostly to you.)
-- "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Re:What else besides games?
by
Junks+Jerzey
·
· Score: 1
Dynamic perhaps, lightweight no.. scripting languages have far more overhead than compiled code, you have to run the interpreter aswell as the actual program, you consume more ram and more cpu every time you run the program, as opposed to just once when you compile it.
But still nothing like compiling a C++ program. I write code in Perl and Python and more esoteric languages all the time. I don't have performance problems as maddening as waiting for a C++ program to link. Ever.
Re:What else besides games?
by
ThomaMelas
·
· Score: 1
And it's not just doing those things individually, but all together. Having your PC record your HDTV like a Tivo, while you do some video editing of your kid's birthday party with good voice recognition.
For the near future, there is demand for faster chips. And having more power allows for idea's that just weren't possible a few years ago. The company I work for does a CCTV DVR product. We can with great ease make a P4 3.2 drop to it's knees and beg for a day off. Having more power would be very, very welcomed by us.
Just wanted to add one thing for those interested enough in science-type stuff that have read this far:
I modeled polymer chain growth in a Computational Physics class using the Self-Avoiding Random Walk method (google it). A 2D calculation of 500000 walkers taking at most 800 steps (usually got stuck after 40 or 50 steps, never got past 600) took an hour or two. The 3D calculation (100000 walkers taking at most 2500 steps) took over a week to complete and I barely got the project in on time. If you are interested in more details, post back and I'll e-mail about it.
Long story short, processing speed matters. Although in this case the 3D calculation was bogged down a lot because it was memory-hungry.
Agreed. The thing that duallie folks call "creamy smoothness":)
www.2cpu.com's forum is a good place to get user experiences. (This to grandparent.)
Re:What else besides games?
by
Dracolytch
·
· Score: 1
While getting my master's degree in Computer Science, I was able to make my P4 whimper for mercy a few times... Honestly, it's to be expected since I was doing things like cross-breeding multimedia files to grow new and interesting files.
These days I'm doing OCR and some interesting image-manipulation stuff, so the faster she runs, the happier I am.
Honestly though, if you're doing they day-to-day of a non-savvy user (Web surfing, e-mail, word), a machine as old as 5 years old could probably meet their requirements. Budget computers are a GOOD thing for low-intensity/entry-level users.
~D
-- This sig has been enciphered with a one-time pad. It could say almost anything.
Re:What else besides games?
by
hmccabe
·
· Score: 1
I run ProTools LE on a dual processor PowerMac G4. When I got the system, ProTools was on version 5 and Mac OS was on 9.1, and it only supported 1 of my processors. When PT 6 for OS X came out, I could finally get dual processor performance. It really does make a difference. I get about 50% more plug-ins on a session before it gives up on me.
Sensationlist statement
by
AKAImBatman
·
· Score: 4, Insightful
He doesn't really appear to offer any substantial concepts for performance improvements. Shrinking the die and upping the clock speed are the most common performance improvements because they are the most effective. Changes to the chips structure or internal coding only result in a one time 10-20% performance boost. And concepts like programmable gateways still have to follow the laws of physics.
Sure, you may be able to optimize a few very common pathways. But you simply can't optimize all of them. Thus a "perfect" algorithm for pathway adaption would again net you one of those 10-20% increases on a general processor. A dedicated machine (e.g. One attempting to calculate PI to infinity) could of course see several times the performance, but then you have to weigh an expensive programmable chip against a cheap custom chip.
Re:Sensationlist statement
by
DigitalDreg
·
· Score: 4, Insightful
You missed the point.
He is saying that people have run out of the easy optimizations. That it is more important now to concentrate on the performance of the whole package, not just the core.
To that end people providing their own macro designs will allow Power to extend in ways IBM isn't planning on. Need better I/O handling? Somebody might sell it to you. Need a cache controller that handles a high number of outstanding cache requests because your software isn't cache friendly? Somebody might have that too. Need to find these people with these designs? They'll all be talking to each other as part of PowerPC consortium...
This opens up avenues for more creative uses that compliement the basic core, and helps bring down design time. Before you might have not even contemplated a custom chip based on a PowerPC design. In a few years, you might be able to glue a few building blocks together to get it.
Re:Sensationlist statement
by
frinkster
·
· Score: 3, Interesting
I had a tough time translating his statements from management/consultant to english, but I think one of the things he was trying to say is that the cost and effort required to continually shrink the die and up the clock rate are growing quickly, so much so that IBM doesn't feel that it's worth it to focus on that aspect in the quest for improved performance.
IBM is surely going to continue to shrink the die and increase clock rate, but it seems as though for the same amount of R&D they feel that there are more gains in performance to be had by looking elsewhere.
Re:Sensationlist statement
by
AKAImBatman
·
· Score: 1
It still sounds sensationalist. Sun has had the Sparc processor as an open standard for years. MIPS is an extremely popular chip for customizing. Despite the thousands of different companies who've customized these chips, neither one has seen a significant divergence in design. A third party will usually add a few instructions specific to their device (e.g. SIMD-like instructions were added to MIPS for video games) and leave it at that.
What history has shown is that general computing is general computing (with an occasional lean toward a certain area such as Desktop processing, number crunching, or network serving). Non-general applications such as GPUs and signal processors tend to require a much different architecture than a general CPU can provide. Thus your car radio is probably processing more data than the PowerBook on your lap, but your PowerBook is running programs that would cause your radio's DSP to choke.
Re:Sensationlist statement
by
Arker
·
· Score: 4, Interesting
I think you grok this well.
Clock speed has never been the main factor in the performance of your computer - it's just been a number that works well for marketing. Your typical modern cpu is idle most of the time anyway. When you increase the clock speed, it does increase performance, but not linearly - doubling the clock speed on your chip might only give you a 10% boost or so in terms of real world performance.
I remember back when the Pentium first came out, having two systems with P60s to compare, the only difference between them being that one had 4 times the cache memory onboard and, I believe, a better cache-logic implementation on board. The system with the superior motherboard was in a whole higher class, performance wise, in regards to every task we threw at it, although the effect was much more pronounced on some tasks than others, it was striking in every case.
As CPU power has been growing far faster than IO capabilities, I would expect the same sort of testing with new systems today would show even more dramatic effects.
Better IO handling is very important for many different applications. Just look at the difference between running an application that will fit in cache against one that requires constant work with your main RAM bank. It's huge. So is the difference between a program that will fit in main RAM and one that requires page swapping with VM. Massive difference. Increasing clock speed shaves a microsecond off here or there, but it does nothing about all the wasted cycles while the CPU waits on IO.
CPU speed over the past 20 years has increased incredibly, but IO capabilities in the PC haven't improved at anything like the same rate. Making CPUs smarter (not necessarily faster, but more efficient at using the speed they already have,) using bigger better designed caches and improving IO systems are likely to be much more efficient ways of increasing real world performance than cranking up the clock speed.
-- =-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
My Asynchronous Processor
by
Throtex
·
· Score: 5, Funny
... doesn't even have a clock, you insensitive clod!
Re:My Asynchronous Processor
by
The+Conductor
·
· Score: 1
No joke, there actually is such a thing. Asynchronous design is much harder to verify than ordinary clock-driven designs, but there is some evidence that that approach may help with current problems with clock distribution (obviously) and power dissipation.
Clock speed long ago stopped being the benchmark of choice for speed. One should look much more carefully at the MIPS or BIPS rating (Millions/Billions of Instructions Per Second).
Clock speed DOES have more meaning for RISC processors. You can judge their performance more or less by the change in clock speed from a prior chip with the same instruction set. However modern RISC processors are not entirely RISC, they contain coprocessors which have instructions which definitely take more than one cycle to complete. Granted the instruction that ships the data off to them should still only require one instruction, and later it should only take one instruction to get it back, so it becomes just a matter of redefinition of terms.
Of course, differences in parallelism, fetch and store speed and so on will make a big difference. The number of functional units and how rapidly you can actually stuff them, the penalty for a bad branch prediction on processors with OoO execution and other factors still have just as much to do with performance as clock rate.
-- "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Re:RISC
by
Anonymous Coward
·
· Score: 0
i would think the clock speed has more meaning for RISC processors
You don't say why you think this is true. I'm going to guess it's because risc chips used to run one instruction per clock cycle.
The reasons clock speed doesn't matter for risc, are the same reasons it doens't matter for cisc. Current risc chips execute more than one instruction in a single cycle, but often stall waiting for cache and other things. With todays high clock speeds, it's the delays that kill performance. The easiest delay to understand is cache misses. Take a look at the performance of identical chips with different cache sizes on the kind of things you do.
<BadAnalogy cpu=engine>. It's foolish to assume the volume of an engine defines which car is fastest, without looking at the whole car. This is true, even if the engine is a diesel.</BadAnalogy>
Except then you have to be very careful about exactly what sort of instruction you are comparing. A chip that does 1,000,000GIPS is not very useful if the instructions are no-ops. One of the reasons that the PPC970 (G5) does well in benchmarks is that it has a fused multiply-add instruction, which is used by a lot of programs. Software would run slower on a machine that executed the same number of instructions per second but did not have this instruction.
If you really want to compare computers, always look at application benchmarks. If the G5 runs photoshop faster, and your main resource intensive task for the machine is running Photoshop, then get a G5. If you are going to be using it to play games then get an Athlon64, or a P4EE (whichever runs the games you use faster). Different processor architectures are better at different things, and so any kind of general speed figure is going to be meaningless.
The x86 chips crack the CISC instructions into RISC-like micro-ops in hardware....so it burns a couple million transistors and lengthens the pipe. Big deal.
Hence there is really no major difference between a "pure" RISC chip such as Power, and a IA-32 chip such as Athlon. Both are superscaler, out-of-order processors.
Itanium's EPIC instruction set you say? Well that was just a bad idea and solves many of the problems from the 80s. The ALAT has been studied and shown to be useless. The problem nowadays is memory latency. A tricked-out ISA will not help much with memory latency on a cache miss. However very large caches do, which is why Itanic does well on SPEC benchmarks. This is also why Itanic costs so darn much as the yields are probably terrible due to the huge chip. Put a 8mb L3 cache on a Xeon and see it outperform Itanium. Of course Intel won't do that for cost and saving-face reasons.
The number of instructions per second is not useful when comparing RISC and CISC architectures. You can do more with one CISC instruction than with one RISC instruction. It is very difficult to compare speeds across architectures except by using common benchmarks.
-- Rhymes that keep their secrets will unfold behind the clouds.There upon the rainbow is the answer to a neverending story
While clock speed is, on the surface, perhaps more relevant to RISC processors, it is not the end-all, be-all metric. Multiple pipelines (particularly specialty pipelines, a la FPUs), exception/branch handling capability (particularly pipeline flushing operations), memory/cache management, and a bevvy of other things are the real drivers for performance.
-- Thinking of starting a business in Minnesota? Me too! mnsmall.biz
AMD has been using RISC-like architecture on their chips since the K-5. Remember the K-5? The K6 lineage expanded on this, as did the Athlon.
I'm not an AMD fangirl or anything but they've been at the "interpret x86 instructions with RISC architecture" game for coming up on a decade. Intel only caught on relatively recently with Banias and Dothan (aka Pentium-M, "Centrino") and are finding out the uncomfortable truth that these chips that operate at a slower clock speed are more efficient and do more per clock cycle than P4s.
And as per Itanium: there's a reason why people call it the Itanic.
-- Knowledge is power. Knowledge shared is power multiplied.
Re:RISC
by
Anonymous Coward
·
· Score: 0
You don't even know what RISC means. It has nothing to do with instructions completing in one clock cycle. In fact, one of the reasons RISC got hyped way back when was that it was straightforward to pipeline the instructions.
Same old
by
Anonymous Coward
·
· Score: 5, Informative
Sounds like the same thing AMD has been trying to convince people of for the past 5 years, while Intel has been lengthening their processor pipelines to ramp up clockspeed while effectively lowering instructions per clock. Unfortunately no one bought it when AMD was saying it, so they had to come out with their PR naming system. Let's hope that at least IBM and their significantly bigger clout can change the picture. It seems like Intel's getting on board too, it seems there are rumblings of them moving their notebook M processors to the desktop as things have gone to hell when transitioning to 90nm fabrication. (In terms of power dissipation)
IBM has clout no doubt, but it will be critical to see how the new G5 Macs perform, as they are the most publically visible implementation of the IBM processor line.
Re:Same old
by
Anonymous Coward
·
· Score: 0
This guy is talking about the Power CPU market, which is completely different from the AMD x86 line of CPU market. They are aiming at banks (first thing that should pop up in your head when someone speaks about "transactions"), big firms, etc. Their clients are well aware of the difference between a CPU speed and a system efficiency, but us mere mortals don't, so he needs to point it out for us. The real interesting twist in this article, is that IBM seems to follow the path of OSS, but now tries to apply the formula to Hardware. Again, being a simple PC customer, I'm not sure if this is really news for the big players (when large sums of money are at stake, I'm sure these guys can find compromises we wouldn't dare dreaming of), since I didn't look for the terms of access to the Power Core "specs".
Seems IBM is embracing open standards
by
foidulus
·
· Score: 3, Insightful
in both hardware and software. While their plan doesn't seem nearly as open as GPL software, it's still a step in the right direction.
If they succeed it doesn't bode well for the x86 architecture, which seems to be a victim of it's own success. They seem to be trapped into just adding faster clocks instead of changing the architecture. They still have neat things like Centrino, but the marketing droids seem to have control over the engineers there. Every update just seems to be a faster clock speed without regards to how much it actually increases performance(I think this is evident in a lot of consumer pc's were they put in the latest and greatest pentium processor but then add in a paltry amount of RAM) I'm not saying I know more than the Intel engineers, I think they are doing a fabulous job with what they have to work with, but...I don't know where I am going with this, I'll just sit back and burn some karma now....damn ADD
Re:Seems IBM is embracing open standards
by
Jameth
·
· Score: 1
"If they succeed it doesn't bode well for the x86 architecture, which seems to be a victim of it's own success. They seem to be trapped into just adding faster clocks instead of changing the architecture."
It seems you haven't been introduced yet. Foidulus, this is x86-64. Him and the opteron family want to have a little talk with you.
Re:Seems IBM is embracing open standards
by
AKAImBatman
·
· Score: 2, Informative
Seems IBM is embracing open standards
It also seems that IBM is a few years late in that respect. (See: IEEE Standard 1754-1994)
If they succeed it doesn't bode well for the x86 architecture, which seems to be a victim of it's own success. They seem to be trapped into just adding faster clocks instead of changing the architecture.
As much as I can't believe I'm defending the Intel architecture, Intel *has* been modifying their chip design. Out of order instructions, Superscalar execution, instruction pipelining, branch prediction, etc. are all in the current classes of Pentium processors. There are two reasons, however, that these don't affect Intel processors as much as other architectures:
1. The awful Intel instruction set was created as part of a quick "stop-gap" product called the 8088 (and later the 8086). Intel had planned to redesign the thing, but got trapped when IBM used it for PCs. This has led to more crappy instructions being bolted on for 32 bit support. This instruction set has made it somewhat difficult for Intel to use traditional performance enhancements. (e.g. A common block of instructions don't quite fit the superscalar block, resulting in lost performance gains.)
2. Intel tunes their chips for video games. Sorry folks, you've all been asking for single threaded performance. Well, single threaded performance is what you've got. The Intel chips will outdo every other architecture on Quake III. Just start praying that your servers don't need to multitask more than 20 or so concurrent threads.
Re:Seems IBM is embracing open standards
by
drinkypoo
·
· Score: 4, Insightful
The x86 instruction set is obviously not the future for desktop systems, at least not in the form of x86. In the near term it looks like that future is x86-64, which is not really the same as the x86 instruction set (which itself has changed over time) though it is dramatically similar of course.
x86 processors have managed to bump the clock and improve the architecture. You have to do both to be successful. Having higher clock rates IS a benefit even if you do nothing else, so long as the rest of the system can keep up. IMO you can give most of the credit for the improvement of x86 to AMD, which really pushed its limits admirably. They also didn't manage to push clock rates as far (at least, not as soon) so they had to add more functional units and tie them all in, which is exactly what you're talking about, an alternative to increasing cycles per second.
-- "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Re:Seems IBM is embracing open standards
by
SoopahMan
·
· Score: 5, Interesting
I disagree - you're forgetting AMD. Intel has been pushing clock speed for a long time, and many consumers are still fooled by this engineered-for-marketing strategy.
But AMD very-much uses the x86 architecture, and has long emphasized things other than clock speed. They've already put into action several of the things IBM's Bernie Meyerson seems to think he brilliantly came up with:
Efficiency: Athlons just plain get more performance per clock than an Intel. There are a lot of factors that contribute to this including the length of the pipeline, but the design just gets more done with each tick. That's less complexity and less...
Power usage: Athlons have 10-12 pipeline stages compared to the Pentium 4's 30. Between that increased efficiency, and less need for a large cache (big pipeline means frequent cache hits), it can use far less power than a P4 for the same performance - and consequently generate much less heat.
Interacting with software: Also not new - more recent desktop AMD chips internally clock themselves up and down depending on whether you're idling or running an app. Laptop chips have done this for years. That means the invalid assumption PC novices make that leaving a PC on while they eat lunch will not use much power becomes valid. For the power user, the PC churns out less heat overall since it only pumps heat under peak usage.
There are things the guy lists that are just freakin' out there:
We are even building in the capability for the chip to physically morph, if required. For instance, you spot an excessive number of fails occurring in the memory--we have techniques in software that recognize those errors. But if it turns out that for whatever reason, one segment of the chip drives an extreme amount of correction, one can easily envision the system autonomically issuing a command to remove that segment.
Uh, dude, this isn't an episode of Transformers, it's a CPU. AMD and Intel already resolved this issue by building very strong chips that don't fail. Even if physically modifying the chip to lop off the bad parts is possible, I can only see it leading to a reduction in quality of chips produced, with manufacturers knowing that worst case, if it fails, it'll just lop itself to pieces.
Re:Seems IBM is embracing open standards
by
ultrabot
·
· Score: 1
(Referring to SPARC:)
It also seems that IBM is a few years late in that respect. (See: IEEE Standard 1754-1994)
But isn't SPARC (the standardized part) only an instruction set? IBM appears to be opening up much more, revealing details about the actual implementation of the architecture.
What I still can't tell is, will it be possible for a guy with a fab to retool his factory for building IBM CPU's, selling them indefinitely w/o paying the license.
Now *that* would be Open Hardware. Being able to improve the architecture is less interesting than allowing various Asian countries (like China) to switch to Power architecture completely. This would be a very efficient way to combat piratism too in countries where it is a problem - none of Microsoft / closed source software runs on the CPU, so there is nothing to pirate.
-- Save your wrists today - switch to Dvorak
Re:Seems IBM is embracing open standards
by
AKAImBatman
·
· Score: 1
But isn't SPARC (the standardized part) only an instruction set?
Nope. The whole chip is available. Sun used to have a PDF of it available for download, but that may have gone the way of the "Free Solaris Source" program.
MIPS is also an architecture that is commonly licensed by third parties. It's used everywhere from set-top boxes, to video game machines, all the way up to SGI IRIX machines. In fact, the many incarnations of MIPS probably made MIPS the *most* successful processor in history.
Re:Seems IBM is embracing open standards
by
harrkev
·
· Score: 1
Uh, dude, this isn't an episode of Transformers, it's a CPU. AMD and Intel already resolved this issue by building very strong chips that don't fail. Even if physically modifying the chip to lop off the bad parts is possible, I can only see it leading to a reduction in quality of chips produced, with manufacturers knowing that worst case, if it fails, it'll just lop itself to pieces.
I think that he is not talking about just re-mapping defects, but actually having the architecture change itself on the fly as needed in order to increase efficiency (but I could be wrong).
While this is cool in concept, I do not even see a concept anywhere. Just throwing the word "morph" around is easy. Telling somebody HOW it is done is an entirely different matter. I am an engineer, so I think that an ideal like this has potential *IF* they can pull it off. But the devil is in the details. I will remain doubtful untill I see some serious nuts'n'bolts discussion on how this is done.
Time will tell
-- "-1 Troll" is the apparently the same as "-1 I disagree with you."
Re:Seems IBM is embracing open standards
by
Dr.+Zowie
·
· Score: 1
Uh, dude, this isn't an episode of Transformers, it's a CPU. AMD and Intel already resolved this issue by building very strong chips that don't fail. Even if physically modifying the chip to lop off the bad parts is possible, I can only see it leading to a reduction in quality of chips produced, with manufacturers knowing that worst case, if it fails, it'll just lop itself to pieces.
That is actually an acknowledged way to boost chip yields. Yields of "good" chips are down around the 10-30% level for the most complex chips (e.g. CPUs) -- the manufacturer throws away something like 3-10 chips for every saleable one they make. That number gets worse as you add transistors: with around 200 million transistors on a Pentium 4, a one-in-a-million failure rate would put 200 bad gates on each chip! One way around that problem is to add redundant copies of key systems and then tell the chip itself to choose the more reliable copy. Then yield goes up because you can
tolerate a few bad or weak devices on each chip.
Re:Seems IBM is embracing open standards
by
SoopahMan
·
· Score: 1
I can see the environmental argument, but you're ultimately agreeing with what I said: You're giving the justification IBM will use to ship lower-quality CPUs to me, the consumer - "it's better than throwing them away and hey, it fixes itself just trust it."
Besides - Athlon chips have shown that lower complexity can result in better performance. If Intel has to throw away more than half their super complex, 30-stage CPUs, I'd say they made their own grave rather than being victims of transistor quality.
Here's an even better answer: Sony's new Cell structure. You make a lot of lower-powered CPUs that can work together on instructions. In terms of the FAB, the much simpler CPUs end up with a much lower fail rate, and the ones that do fail are sufficiently small and low-cost that total loss is minimal.
Re:Seems IBM is embracing open standards
by
Dr.+Zowie
·
· Score: 1
I suppose we are in technical agreement -- I just don't think that a chip with planned dead zones on it is a lower quality product: it might take slightly more real estate than a "perfect" chip, but the size of the CPU itself is not a major headache -- the package size is determined by the pin count. So long as the circuits that I'm actually using are up to spec, the product is "perfect". But then, I'm no overclocker...
Re:Seems IBM is embracing open standards
by
Raptor+CK
·
· Score: 1
It's not a CPU, it's a mainframe processor. I'm sure they're made to even *stricter* tolerances than a desktop chip. However, chips fail. It's inevitable. What they're doing here is engineering a radically more complex CPU (with the associated increase in cost) in order to make a CPU which will route around failure. Sound familiar?
If part of the CPU fails in a live system, and you can't afford any downtime, it's better that the system reroutes around the failed portion of the chip, and alerts the sysadmin. Either way, you're going to need to replace parts at some point. This just reduces the downtime between replacements.
-- Raptor
"Procrastination is great. It gives me a lot more time to do things that I'm never going to do."
Re:Seems IBM is embracing open standards
by
psetzer
·
· Score: 1
The reason that there isn't much improvement on the 80x86 architecture is that it's about as tricked out as it's going to get. I mean, a +7 blessed rubber hose is nice and all, but there's no room for improvement, but the -5 cursed Greyswandir, while inferior to the hose at the moment, has a huge amount of potential. Frankly, the 80x86 architecture is probably one of the last legacy systems I would ever want to build on top of. MIPS is much nicer except for the inability to JMP across certain memory boundaries.
-- "Anyone who attempts to generate random numbers by deterministic means is living in a state of sin." -- John von Neumann
thanks for the link
by
mattkime
·
· Score: 4, Funny
to IBM's website.
I hate it when articles are posted about small, obscure companies and then I can't find their website.
-- Know what I like about atheists? I've yet to meet one that believes God is on their side.
Re:thanks for the link
by
dasmegabyte
·
· Score: 1
I hope it's fast enough for you. I hear the webserver, like most of the rest of the company, got farmed to India.
Depends on your interpretation...
by
Mz6
·
· Score: 1
Does average Joe mean your average web surfer? Only online to pay bills, surf the internet, read mail and send pictures. Or is your average joe the person who uses it for editing home videos plus the average internet user.
I think you might be seeing a plateau on "average" users purchasing bigger, better processors or computers. If the computer they have can get the job done for them and they aren;t moving on to other computer tasks (ala video editing, rendering, etc..) they don't mind having 2+ year old computers.
-- Hmmm.
Re:Depends on your interpretation...
by
Paulrothrock
·
· Score: 1
That is, until Longhorn comes out and the default desktop background is a real-time rendering of a fuzzy bunny jumping around. Then your two-year-old computer won't seem so hot!
In other words, for Internet and email, Linux is fine.
When I buy a PC I like to base it on figures and what it can do (AKA can it run UK 2k4, will it run HL2 for example). Rather then going "ooohhh pretty case!"
I'm funny like that, wonder why..
-- ---
[Insert intresting Sig here]
It's all marketing
by
grunt107
·
· Score: 5, Insightful
The processor speed for marketers is comparable to the engine size wars in the 60s/70s. If I say I have a 402 (6.6L) in my Chevelle and Bob next door has a (snicker) 350 (5.7L) in his Nova, my car gets the approving nods, but may not be faster since the Nova is lighter. Now compare said Chevelle w/today's Z06 'vette. Little 'wimpy' vette has just a 5.7L, but kicks the snot outta the Chevelle in performance.
IBM, and other marketing 'geniuses', need to name their products to entice the 'mine is bigger' crowd. Right now, in the consumer computer realm, GHz talks. Most non-IT people I know will spout the "My PC is 4GHz - what's yours?" mantra when a 2.8 Opteron w/SCSI320 will kick its butt. The enlightened will know, but 'tis the general ignorant masses that have the buying power.
Re:It's all marketing
by
pknoll
·
· Score: 4, Insightful
To extend your car analogy a bit further -
Both a Kenworth over-the-road tractor and a Formula 1 car have about 1000 horsepower. But one will accelerate a LOT faster than the other. And one can tow 20 tons of stuff behind it.
Even IF MHz were directly comparable, you still couldn't judge the speed of a computer without considering what that computer was built to do.
very nice analogy. i would mod you up had i the points to give. speaking of SCSI320... hard disk is one crucial factor in performance to which most people don't pay attention. i've only just come to appreciate it myself, doing my first work with some (relatively) large production databases.
even on my workstation on the job, my P4 2.8GHz HT processor regularly waits and waits for the puny IDE HD to load or seek through a data file or complete a search of the filesystem. it's like wasting access to a genius: he could solve any problem within seconds, but it takes an hour to ask the question.
Both a Kenworth over-the-road tractor and a Formula 1 car have about 1000 horsepower.
While I won't disagree with the F1 car (although you never can tell since F1 teams tend to keep exact HP figures a secret), I doubt you'll find many Kenworths (or any other trucks, for that matter) with 1000HP.
If you want to look at the two largest truck-engine manufacturers, Caterpillar is currently producing a 625HP engine (the C-15), while Cummins is producing a 565HP engine (the ISX). In general, at the speeds travelled in North America, combined with load restrictions imposed by law, that's plenty of power for most drivers.
Sorry to be OT, but I just thought I'd point this out -- my father was a truck driver for many years.
What's a 2.8 Opteron and how does SCSI320 imply the speed of the disk? Are there any disks that saturate a SCSI320 bus? I think not. Disk is irrelevent when you are waiting on the CPU anyway. And just how does one computer kick another's butt? There are various levels of enlightment apparently.
But it won't be long until everyone laughs at the idiot sliding off the track in his car made out of railway tracks and fibreglass.
Brand loyalty can go a long way as when a lowered Honda gets passed by a hot Volvo but the Volvo gets laughed at when everyone's doing laps of a Friday evening.
"RISC architecture is going to change everything"
-- *#*#*#*#*#*******
I love peanut butter sandwiches!
> The processor speed for marketers is comparable to the engine size wars in the 60s/70s.
When I am talking to non-techos I draw the analogy megahertz to processing power is like revs per minute to horsepower or torque. Intel CPUs have high RPMs in the tacho but less horsepower and torque.
Re:It like guys with short penises...
by
Anonymous Coward
·
· Score: 0
Isn't that 'penii'?
Sorry, couldn't help...
Make the chips run cooler
by
Hoonis
·
· Score: 4, Insightful
Corporate datacenters are now filling up with half-full racks because cooling & power requirements are through the roof. You end up being unable to increase compute resources because you have to put in fewer of the faster systems, gaining you nothing.
So hey, if you're listening harware vendors, see if you can't simply make the dang things run cooler on less power before you speed them up!
Apple has been trying to get this message out for a while. We had a story a few months ago (lazy me, no link for you) about how Intel was dropping the clock from the branding of the processors.
Clock speed really does not have a direct correlation to computer speed anymore. It seems like we will see more of the trend of newer, better technology that runs at a lower rate but executes a lot more in one tick, so it is much faster. It seems that it will start at 1GHz and move up to 3 before somebody gets a new idea, makes a new "slow" processor and starts it over...
Actually, it does. Just not across different processor lines.
Apple's message has actually been "don't pay attention to clock speeds since we suck at that". That's part of their problem of "we're stuck using a decade-old POS G4 while everyone else has a modern processor". Now that they have a modern processor themselves in the G5 they're much more agressive with their performance claims.
Once Intel bails on their crappy P4 concept clock rates will largely normalize. Pentium-M, Athlon64 and G5 all have clock rates that are more alike than different. That will be the future once the P4 is euthanized. The whole "clock rate doesn't matter" was invented by Apple as an excuse for being hopelessly behind and is now propped up by Intel and their stupid 30-stage pipelines. Clock rates do matter. There's only two ways to go faster---get more done per clock or get more clocks per second. Designers try to do both and marketers produce BS.
GHz is the wrong metric
by
kennykb
·
· Score: 5, Insightful
Most applications nowadays founder on memory
hierarchy performance (L1/L2 cache, main store,
backing store). Cache misses are a usual killer,
and fetch prediction doesn't work very well at
all yet.
Even on the base CPU, the most important metric,
I find, is "MIPS per watt". That's what determines
how much horsepower you can get off a given amount
of cooling, which is the real limiting factor for
CPU speed.
Re:Mirror , just in case
by
Anonymous Coward
·
· Score: 2, Funny
Mirroring ibm.com should probably be modded Redundant:D
*Everyone* knows?
by
funkdid
·
· Score: 5, Interesting
I think to the/. crowd this is certainly old news. Ever try to explain this to grandma? Or your girlfriend's little brother? *Most* people after my speach of how processors work say "Yeah but arent AMD chips slow? Like a Pentium is 3Ghz, AMDs are cheap (meaning cheaply made) right?"
So I "dumb" my speach down a bit and give it again. The masses don't want to know how processors work, they don't want to know about architecture, they want an even base line to measure performance. Most people think the Comp Usa rep is ripping them off and they are trying to feel good about their purchase, being an un-educated consumer.
By buying the high clock speed they can compare it to their neighboors and in their heads they have a Super-Fast PC.
I'd like to note that most people I talk to look at AMD like most people look at a Yugo. (remember those cars?) In spite of my advice that an AMD is like a new Honda for $2,000.
Rather than trying to explain the nuts and bolts of why clock rate isn't important, perhaps you should try an analogy. I haven't actually had to explain it to anybody, but I figure that I would compare it to trying to determine the speed of a car based by reading the tachometer. In such a case, the gearing compares to things like IPC of the processor.
-- I've got a mind like a steel trap - it's got an animal's foot stuck in it.
Re:*Everyone* knows?
by
dknight
·
· Score: 2, Funny
I love AMD man, but honda? I think you need to find a SLIGHTLY better car to compare it to;)
How about Mazda?
(yea, ok, fine, I'm one of those people that wont touch a honda - except a prelude SH or an S2000 - with a 10ft pole)
When I ask these kinds of questions people usually roll their eyes and tell me to F# off; sociable types of people....
I don't know that much about processor design or performance trade offs, though I know the generally important terms are RISC CISC, L1/L2/L3 caches, FSB and clock speed.
Can you point me to an online resource that presents another level of detail without drowning me in too many orders of magnitude of detail?
Thanks, -S
Digital Signal Processing
by
Detritus
·
· Score: 1, Informative
There are many applications, like software defined radios and televisions, that require huge amounts of number crunching.
-- Mea navis aericumbens anguillis abundat
Re:Digital Signal Processing
by
the_2nd_coming
·
· Score: 0, Troll
and a person using a run of the mill cpu for DSP is a fucking moron
--
I am the Alpha and the Omega-3
Check your cooling
by
Trigun
·
· Score: 2, Informative
The processor shouldn't just lock up like that. Perhaps the heavy load is overheating it?
Next time I'm going to start about this priest that enters this bar and...
And just wait for the +5 Interesting to come.
-- O make me a mask
Re:Bogo Mips
by
Anonymous Coward
·
· Score: 0
BOGO is an acronym for "buy one get one [free]" - typically used in promotions at various stores.
Ah, you gotta remember--half these people don't even go to stores. I myself prefer to nibble on the little dust specs that accumulate on the bottom of my case.
Re:Bogo Mips
by
Anonymous Coward
·
· Score: 0
No, it should be moderated insightful. Slashdot moderation itself is a joke and should not be taken seriously.
I contribute my spare computing cycles to distributed computing efforts. The distributed.net client, at least, does not require the computer to be switched on 24/7 either. Even playing DVDs only take up about 25% of the CPU time, and I've had no problems with overheating. I sometimes think about all that A/C power and computer cycles being wasted at the university computer rooms.
Maybe O.S. vendors could include a voluntary option doing install to contribute your computing power. They could then re-sell that computing to users who need to do heavy calculations that are not time-critical or secret.
P.S. IBM is hot on grid computing, but they don't all seem to believe the processing power also can be contributed by individuals.
says processor chip speed is old news
Half the posts on/. already say this... I saw two different posts today from different topic headers, one was from the Apple post on the G5
--
I mod down so you can mod up. Your welcome.
Transcoding compressed video files...
by
FatSean
·
· Score: 0
Converting a bunch of AVI to VOB format so I can burn them takes like 9 hours for 80 minutes of resulting video on my 1GHz P3. A 1.6Ghz P4(it's in a laptop w/ speedstep...not sure of name) does it in 3.5 hours.
-- Blar.
Uh, can you spell AMD
by
Omega1045
·
· Score: 4, Interesting
Didn't AMD take this approach some years back? They have to name their processors to sounds like pentium clock speed ratings, but they have been preaching the idea that clock-speed is not the sole issue for years. I know IBM is technical leader, but it just smells like IBM, like Intel, are jumping on the AMD bandwagon, but they aren't calling it the AMD bandwagon.
--
Great ideas often receive violent opposition from mediocre minds. - Albert Einstein
Everyone knows....
by
Anonymous Coward
·
· Score: 0
that its not the clock speed that matters, but how high the number is beside the processor that matters....duh.
would want there clock to go at more than (1/60)Hz? Im trying to get MORE SLEEP, not less. Geeze, just increasing clock speed to 1 Hz, means that in a typically day, I would get.1333 hours of sleep on a good day.
Re:who in the world
by
Anonymous Coward
·
· Score: 0
Idiot. Hz means cycles per second. 1/60 Hz is a cycle every minute.
"...because if you let the smoke out, it stops working."
- my friend, when he's asked what's inside a transistor.
-- SIGERR: laziness exceeds quota
a stick is a club...
by
simpl3x
·
· Score: 3, Insightful
business is often about defining your strategy for approaching new business. ibm is stating that openess will benefit their business. others have recognized this holistic approach to systems design, which from the beginning of computing was really required for high end/specialized systems. some wanted openess for additional reasons, such as freedom in beer and speech. so the same stick has simply been picked up to use in a more competetive environment against businesses not capable of integrating both sides of the computing environment.
so long as everybody has their needs met, it's a good thing(tm).
what I want to see
by
AviLazar
·
· Score: 3, Insightful
I don't care if they want to name it by processing speed, bus speed, or hell how much donkey speed it has - i just want it to be consistent!!!
That means give me Intel processor 1, then the future processor named Intel Processor 2 WILL be better then the first one, and the third one will be better, etc.
Stop coming out with funky new names to confuse me (and the less informed computer users, which i am probably in that category).
If I want to look at the specs I will, but at least make it easy for people to realize which processor is the latest and greatest!
I think the worst case for this is some of the graphics cards - whats better the 9600 version or the FX version or the super version or the crack version?
Come on, have some sympathy on us ----damn marketers!
In other words, the computer is helping to design computers that are faster than itself? But then you go out to design an even better computer? That can then design an even better computer? How ironic it would have been if the only application of computer technology was creating new computers.
"I speak of none but the computer that is to come after me, A computer whose merest operational parameters I am not worthy to calculate. And yet I will design it for you."
-- Irene KHAAAAAAN!
The most direct processor measurement
by
Profane+MuthaFucka
·
· Score: 4, Funny
The old MHz measurement was nice for a time, but it just wasn't measuring what computer performance geeks find important. Just like car performance enthusiasts like to talk about the horsepower ratings of their engines, what is really important is a measurement of how the system really performs. That's why a quarter-mile timing can be more informative about a car's performance than just looking at the car's horsepower and torque rating.
So, the new processor measurement that gets right to the heart of what's important for a CPU performance geek is going to be henceforth the PU. The PU stands for 'penis unit' and it indicates to their fellow cpu performance geeks how big their dick is relative to everyone else's.
-- Fascism trolls keeping me up every night. When I starts a preachin', he HITS ME WITH HIS REICH!
Power is not for PC
by
rve
·
· Score: 4, Interesting
The POWER architecture isnt really for the average Joe's computer, it is for servers. In servers, many tasks are done by coprocessors and independant subsystems without taxing the CPU. The extra CPU performance is now suddenly needed because IBM keeps encouraging ISV'S to write for Websphere, in Java, so you now need 10 times more memory and CPU performance than you previously did to perform the same task. In servers, the worst bottleneck at the moment are afaik still the moving parts in the disks and tapes.
The PPC is a cousin of the embedded version of the chip, where the performance per watt power usage is relevant. It is hugely successful.
Sales of Apples with desktop POWER chips aren't really significant. Although IBM aren't ready to admit it yet, the battle for the desktop is long over. No amount of performance advantage is going to outweigh the main advantage of the x86 architecture there: backward compatibility, preserving the value of past investments in software for the customer. IBM should know this, as they still make their zSeries mainframes compatible with the 40+ year old 360 architecture for the same reason.
In the PC, unlike most servers, most everything goes through the CPU, which is why for the average Joe raw CPU performance _does_ matter.
Re:Power is not for PC
by
Anonymous Coward
·
· Score: 0
You aren't keeping up with the press releases. The "PowerPC" brand name is going away. (Which is a good thing, because it's rather stupid sounding.)
Future PPC processors, for Apples and embedded, will be branded "POWER" just like the big chips.
Re:Power is not for PC
by
cbiffle
·
· Score: 4, Insightful
The extra CPU performance is now suddenly needed because IBM keeps encouraging ISV'S to write for Websphere, in Java, so you now need 10 times more memory and CPU performance than you previously did to perform the same task.
Your post is, for the most part, dead-on and well-put, but I can tell you're not an enterprise Java developer.
Our transaction processing systems were recently moved to Java from C (Solaris on a Sunfire 6800, 8-way SPARC).
Yes, they require more memory. This doesn't really concern us because we spend far less time tracking down dangling pointers and memory leaks now. The increase in memory seems to be about 4x-6x for our system, which still brings it in under a gig.
No, they do not require more CPU. Several parts of our application actually run faster than the C version. I credit the Hotspot on-the-fly optimization crap for this to some degree, but I'm honestly not sure what the deal is. (And I'm our profiling guy. Ain't that sad?:-)
But more importantly, as you mentioned, on big iron the I/O throughput tends to be the bottleneck anyway. Our transaction-processing systems tend to sit happily with significant idle percentages while positively slamming the disks and databases.
We're running inside Sun's Solaris JVM in a hacked-up proprietary version of EJB, using Tomcat for the frontend. I can't imagine that Websphere has much higher overhead, though I could certainly be wrong.
Re:Power is not for PC
by
rve
·
· Score: 3, Informative
It's not the programming language Java perse, but the writing in an interpreted language for an application server sitting on top of a virtual machine sitting on top of the operating system's HAL sitting on top of the hardware as opposed to writing a natively compiled app for the HAL as we did before.
But you're right, I'm not an enterprise Java developer.
Re:Power is not for PC
by
rve
·
· Score: 2, Informative
This doesn't really concern us because we spend far less time tracking down dangling pointers and memory leaks now.
That isnt really an issue for COBOL programmers hehe:)
No, they do not require more CPU. Several parts of our application actually run faster than the C version. I credit the Hotspot on-the-fly optimization crap for this to some degree, but I'm honestly not sure what the deal is. (And I'm our profiling guy. Ain't that sad?:-)
Maybe it's the fact that java has a far better standard library that makes the new application faster. I often see C programmers confusing the C language's potential(!) for higher speeds somehow with their programming skills, and they seem to be suffering alot more from the not-invented-here syndrom.
Not for the customer surely. The investment that is being preserved is for Microsoft and their code. There aren't that many customers that require the ability to run their old MS-DOS applications. Microsoft would rather you buy a new version of your app that runs on Windows. If I really want to, I can run MS-DOS apps on my PPC system.
The company I work for was forced to move off of a DOS application because the vendor didn't want to support it anymore. Guess what, the new app is written in VB. It has no new features and it's GUI is from HELL.
-- It's mandatory to wash your hands before returning to the land of Dairy Queen.
compilation to be a quick as possible.
by
dpilot
·
· Score: 1
In which case you're more interested in the disk and memory subsystems than in raw CPU speed. From what I've heard, VMWare is also just as much a memory hog as CPU hog, too.
For both of these cases, seems to me that your incremental dollar would be better spent on more memory and/or faster FSB instead of a simply-higher clock speed CPU.
Obvious caveats: Once you hold your working set in RAM, there may be little point in more memory. Assuming you're 32bit, the 4G+4G patch may hurt more than it helps, and you may be max'ed in the 1-4G range, but how big is your source tree? Higher FSB and clock speed are somewhat related, but there's still wiggle room on both Intel and AMD.
-- The living have better things to do than to continue hating the dead.
The Article: Distilled
by
Dr.+Smeegee
·
· Score: 2, Funny
Addressing a fundamental shift in the landscape of technology, a significant shift in the trajectory of basic technology. The rate of performance enhancement is becoming impacted as holistic design. You have to have a means by which you proactively and holistically address that extraordinary event. I am hoping that people really understand the sort of discontinuity we are talking about the capability for the chip to physically morph, one can easily envision the system autonomically issuing a command moving off to an entirely different plane.
The world is full of incredibly bright people.
This is why I hate slashdot.
by
Wakko+Warner
·
· Score: 1
What other applications besides games really tax the CPU right now?
Are you planning on a career in the computing field? If so, please send me your name and address info so I'll know never to hire you.
If you think the only thing a fast CPU is used for these days is playing "Unreal Tournament", you have no business managing (or even logging in to) the kinds of UNIX boxes some of us deal with on a daily basis. Yes, a faster CPU (or 24 of them) is more than necessary for some of the applications we run in the Real World.
No, we're not beta-testing the next "Doom". We're building aircraft.
- A.P.
-- "Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
AMD and even Cryix before them have been telling us this for years..
-- You have 5 Moderator Points!
Which Helpless Linux zealot/MS basher do you want to mod down today?
If you can beat 'em, change the rules of the game.
by
stienman
·
· Score: 4, Insightful
IBM is attempting to change the rules of performance measurement. They are doing this by educating their customers. Inherently, people want a single performance metric that says X is better than Y because this perfromance metric says so.
IBM would prefer customers to come to them and ask IBM, "Which processor is better?" rather than rely on an external, easily verifiable, though not accurate, single number indicator.
The truth, as we all know, is that there is no single metric since each processor has strengths and weaknesses and various applications rely on these strengths differently.
They are also opening their processors to the end users a little more, almost as a jab at Intel. Intel has microcode, but you'll never see it or get to modify it. But the very presence of microcode in almost every modern general purpose CPU means that performance can be enhanced and tailored for each application with very little processor change.
So IBM is letting people get closer with the processor to enhance performance with very little risk or effort.
The kicker is that it's not simple, so only a few large manufacturers and some dedicated homebrewers will really have anything to show for it.
Thus it's a marketting ploy intended to raise questions about current performance metrics in the minds of indecisive consumers.
But then, when has the CPU war ever been about anything but marketting?
-Adam
Please RTFA
by
Anonymous Coward
·
· Score: 2, Insightful
This is very significant, and not because of the clock cycle bit. 90% of the comments are responding to the article title, not content. Typical.
The open architecture ecosystem can have a far reaching fundamental effect on potential market-space of CPU dependent products.
This is far bigger news than any of Sun's open sourcing or Brown v. Tanenbaum.
One thing I've learnt over the years is that just because a company talks about open standards doesn't mean that there is.
If I can build my own fab and start producing exact copies of IBM's design and sell them without even seeking permission, then I'll believe it. (same with the sparc) Until then, it's all talk.
drive a Ferrari to the convenience store...
by
cenonce
·
· Score: 1
This is all marketing. Where does having the fastest processor really matter? Gamers, maybe, but what gamers buy a Mac with an IBM chip to play games?
Next would be digital audio/photo/video, and that is all pretty much wrapped up by PowerMacs. You don't need to sell to those professionals and they generally have the money to fork out 4 or 5 grand for a dualie Mac every other year. Afterall, its how they earn their living!
99% of the people surf the web, check e-mail, pay their bills online, write letters and occassionally play a game more complicated than free cell. You can do all that with the slow-assed 733 mhz G4 I have (with no L3 cache to boot)!
These people must also feel the need to drive a Ferrari to the convenience store.
Re:drive a Ferrari to the convenience store...
by
pseudochaotic
·
· Score: 1
733! Pfft! I could do all that on my old 66mhz Pentium!
-- And the l33t shall inherit the 34r7h.
Re:drive a Ferrari to the convenience store...
by
Zcipher
·
· Score: 1
This is all marketing. Where does having the fastest processor really matter? Gamers, maybe, but what gamers buy a Mac with an IBM chip to play games?
Pick me! Pick me! Believe it or not, I buy a new mac based on whether or not my old one can play the games I want it to. Do I wait longer and have less selection? Yes. Do I particularly regret it? Not so much. Would I buy a G5 to play games on? HELL yeah. For now, I'll just have to live with my nice, new iBook, though ^_^ (bought primarily because I desparately wanted a laptop, but also nice because it can play stuff my elderly desktop can't)
Wake me when it's over
by
Anonymous Coward
·
· Score: 0
"...a fundamental shift in the landscape of technology..."
ZZZZZzzzzzzz.....
Re-programmable
by
The+Conductor
·
· Score: 4, Informative
There is a trade-off between speed, reliability, cost, and re-programmability.
SRAM types
Are re-programmable but require a rather slow serial load at boot-up. Reliability in embedded systems leaves something to be deisired since any brownout-induced glitch can create errors that are even worse (harder to recover from) than software glitches because wired logic doesn't have anything equivalent to code checksums or interrupt vectors. Well-paid FPGA designers are versed in the arcane art of self-verifying logic.
EEPROM types
Come alive at boot up and are much more resistant to glitches. Their performance, however, is slow. And you have limited (100,000 maybe) rewrite cycles.
Anti-fuse types
are made by Actel. They have the highest performance and best density. They come alive at boot up and are dead-nuts reliable under the worst of conditions; for example, properly qualified, they can survive the cosmic radiation in spacecraft that would leave other types toasted. The big drawback: the anti-fuse process, which works by melting diodes into short-circuits, is not eraseable.
Desktop systems (say, an add-on FPGA card) would be best served by SRAM types, since you already have a processor that requires gluttinous gobs of puritanically clean DC power. Basement hardware hackers would be better served by EEPROM or anitfuse types (depending on performance requirements), since they don't require super-expensive exotic design software.
Re:Re-programmable
by
harrkev
·
· Score: 3, Insightful
I would disagree with the last statement. Xilinx FPGAs are perfect for an experimentor. They can be easily programmed with a JTAG cable, just like everybody elses parts. And Xilinx has low-cost and free design suites available. This makes is perfect for development/debug. A home experimentor is likely to make a LOT of mistakes when designing, and EEPROM-based parts take longer to program, and they DO wear out after burning too many times.
However, in order to program a Xilinx part in an embedded system (without a PC attached) requires a way to program a serial EEPROM. Programming this might be a pain, but Atmel (for one) makes serial EEPROMS for just this purpose, and will also be happy to sell you a programming cable.
-- "-1 Troll" is the apparently the same as "-1 I disagree with you."
Xilinx has low-cost and free design suites available.
I peeked over at xilinx.com, and <Gomer Pyle Voice>Shazam!</Gomer> They do now. A few years ago they really couldn't because AT&T/Lucent was a pin-compatible 2nd source (AT&T got a license in exchange for fab capacity, I think). So software support couldn't be justified as a sales cost...people could just get the cheaper Orca chips. It appears that the Xilinx 3000 & 4000 series are long gone now, and Xilinx now fabs its chips elsewhere.
There's still glitch control, though. Not a problem if your project is gonna plug into a host system (such as a PC) that has already solved power supply & shielding issues for you, so bring on those home-brew hardware MPEG encoders! But if your project is some battery-powered robotic gizmo, or a digital RF synthesizer, an SRAM based FPGA could be awfully glitchy.
Disclaimer: my wife sells Actel (which makes antifuse & flash types), on commission but over a limited territory. (In a previous job she used to sell Xilinx.) I have used SRAM and anitfuse designs (though I am not really a digital logic guy) but never flash.
Yeah, well tell your wife that is totaly SUX that ProAsic Plus makes you use Design Compiler instead of the more sensible (and easier to use) FPGA Compiler II.
I don't sound a little bitter, do I?;)
-- "-1 Troll" is the apparently the same as "-1 I disagree with you."
Maybe you can generate intermediate VHDL with FPGA Compiler and then compile that with the native Actel tools, though that leaves the timing debugging directly in VHDL, I think. (Supposedly, if you migrate the functional design to an antifuse device, the timing problems simply go away because the parts are so bitchin' fast, but I cannot attest to this personally.) FPGA Complier is a 3rd party tool (by Synopsys) so I don't think there is much that Actel can do about it.
But then again I have never seen an FPGA design tool that I liked. The schematic-based designs are hell to maintain; version creep in the "parts" libraries drove me crazy. VHDL is a bit removed from the solder, kynar wire, and test probe mentality of someone like me, a mostly analog/RF guy. Without a schematic, where do you stick your simulated probe on your simulated circuit? It's bad enough that I can't jab a real probe in there. It's enough to make you think it's all somehow Microsoft's fault (since MS is clearly the source of all evil in the world).
so it burns a couple million transistors and lengthens the pipe. Big deal.
But it is a big deal. First of all, a lengthened pipeline will result in more costly branches. As a result, more transistors will be required to beef up the branch prediction + cache. And all these extra transistors will add to the power consumption and manufacturing cost of the chip. So you end up with a chip that uses far more transistors then an equivalent chip with a smarter instruction set / design (just compare transistor counts between the P4ee, Power, and Athlon64 of the same speed - the extra long pipeline of the P4ee hurts).
Most of the statements you made were dead on and I would generally consider your post to be a good one. But I can't help but imagine what todays CPUs would be like if everyone used an open source OS that could be run on any CPU architecture. This would result in a more competitive market by making it easier for new companies to enter the market with cool new designs. I'm willing to bet that x86 chips would have died off long ago.
What I find interesting is the way in which the Itanium gets around the problem of branches. We're all aware that the majority of conditional branches seem to be for loops or skipping small sections of code. By allowing for conditional execution of code, they can get rid of all the JMP three lines down instructions and make it so that there is less pipeline blocking going on. I'm suprised it isn't being used with all sorts of chips nowadays since it fixes the biggest problem with pipelines.
If you force the programmer to make a guess on if a branch will be followed or not when they write the code, then it's possible to get rid of much of the branch prediction code. In essence, if you put more demands on the compiler writers, then you can make much faster chips without the need for obscene numbers of transistors.
-- "Anyone who attempts to generate random numbers by deterministic means is living in a state of sin." -- John von Neumann
What I find interesting is the way in which the Itanium gets around the problem of branches. We're all aware that the majority of conditional branches seem to be for loops or skipping small sections of code. By allowing for conditional execution of code, they can get rid of all the JMP three lines down instructions and make it so that there is less pipeline blocking going on.
Predication is not a new idea. ARM has been doing it since the 80s.
But it is a big deal. First of all, a lengthened pipeline will result in more costly branches. As a result, more transistors will be required to beef up the branch prediction + cache. And all these extra transistors will add to the power consumption and manufacturing cost of the chip. So you end up with a chip that uses far more transistors then an equivalent chip with a smarter instruction set / design (just compare transistor counts between the P4ee, Power, and Athlon64 of the same speed - the extra long pipeline of the P4ee hurts).
The micro-op cracking uses less than 5% of the transistor budget. Plus the P4 trace cache stores decoded micro-ops (not CISC instructions).
As far as branch mispredictions, you have a point. A longer pipeline can make the penalty greater. However the P4 also uses selective replay to mitigate this. This combined with the effectiveness of the trace cache makes the branch misprediction penalty much smaller than you might think.
This sounds like an ad for an FPGA. But they don't want to make it too flexible:
The architecture, therefore, must be protected, because that is what establishes how you communicate. So, yes, we have to lock down the architecture so that it cannot be randomly disrupted. Otherwise, imagine that somebody came along and altered the core architecture along with the core instruction set. It would invalidate all the prior intellectual property and destroy the ecosystem.
FPGAs are useful, but very few applications reprogram them while in use. I doubt that's what they intend.
Machines have been built into which you could load application-specific microcode. The DEC VAX 11/780 had that feature, although nobody ever used it. Shader programs in GPUs are probably the most successful example of this idea.
They'll probably export some internal tuning parameters, like the instruction/data ratio for the cache or the depth of instruction prefetch, produce some benchmarks showing that if you tune those performance goes up 20% or so, and hype that into "reconfigurability".
thoughts on clockspeed and CPU power/ratings
by
itzdandy
·
· Score: 3, Interesting
here is a thought.
When you buy a car, you don't just consider how fast it goes, you consider fuel economy/comfort/quality etc. this could be applied to CPUs
for instance, you buy a new DeLL/HP/whatever, the machine has 3 numbes on it. 12/16/65 - which means, 12 is the general office app benchmark, 16 is the gaming benchmark, and 65 is the mean power usage of the machine.
so a good office machine is a 14/2/30, but if you are playing games, you need a 6/26/130, you don't care as much about the power bill or how fast office computes a6:c6*c14 whatever. these numbers would be linear, as in the nex-gen would just have higher numbers.
Each computers label would ave a description of the rating above the label saying "look at the killer gamer system" or whatever.
I can see the arguement of the system being confusing, but i'd take the least confusing method that was effective, and i think this would be effective.
------------------
Something like this could translate over to server side with web/fileserving/powerreq or something, but it would allow companies like AMD and IBM who have not pushed the MHz myth to the extreme to allow their product to compete on merrit not Mhz.
thoughts??
Re:thoughts on clockspeed and CPU power/ratings
by
evilviper
·
· Score: 1
When you buy a car, you don't just consider how fast it goes, you consider fuel economy/comfort/quality etc. this could be applied to CPUs
I like the idea, but it's a terrible analogy...
When you are driving down the road, you aren't going at your car's maximum speed all the time... With a computer, you are. If cars could safely go at 500MPH, with no risk of death, injury, or discomfort, you can bet that people will be willing to spend the extra money on gasoline to fuel their 500MPH cars.
Most people want to get things done quicker, and that's what faster, hotter processors offer. They don't think about the fact that they have Microsoft/KDE to blame for their snail-like performance, they just know that a faster processor will make them able to do more in less time, with less waiting.
Personally, what I want to see is not power ratings listed (not that it would hurt), but computations per watt/heat, and even more, I'd like to see min/max power used. In other words, I'd very much like to see CPUs that only use 5watts while completely idle, even if they have to use 100watts when they are at full load. That would result in more electricity savings, and heat savings than running a ~60watt system all the time (not that current systems are 60watts, just an example).
Re:thoughts on clockspeed and CPU power/ratings
by
Qutec
·
· Score: 0
If her specs are 6-26-130, you have more to worry about than processing power,..like food consumption.
Re:thoughts on clockspeed and CPU power/ratings
by
SignificantBit
·
· Score: 1
yep.... the car is the computer, the motor engine is the cpu.
Re:thoughts on clockspeed and CPU power/ratings
by
itzdandy
·
· Score: 1
well, how about this
you need to go 500MPH on a crotch rocket to get 1 person to 1 place in a certain about of time, then you need to make 20 trips to get 20 people. OR you could take a bus that holds 20 people, make 1 trip, and do it at 25MPH, so you could kick that up to 50MPH and move twice as many people in less than half the time.
the whole landscape of computing could shift to more SMP capable systems. Ideally, you should have some interconnect for CPU's that was more modular. Have a high speed interconnect for just processors OR have a universal high speed interconnect for everything including processors. have the system see a CPU simply as somthing that run processes, and have some small memory on the chip that let the system know what that CPU could process and how man MIPS/FLOPS/whatever it was capable of running.
You would boot up your OS and have X amount of processing power, then you could plug in a CPU card into a spare slot and it would hotplug and be added into the processing pool for whatever instructions it was capable of processing.
ideally, you could run multiple intruction sets on any given system, because when you plugged in a new processor, it would allow those new instructions to be executed, your powerpc programs would no longer segfault OR they would not have to run in emulation space.
you could also have the system use power management to shut off extra processors and put their abilities into a "powerup if needed" state, so you would only power up for floating point processor if needed, and only run it at the Mhz needed for the task at hand.
-- Nothing in the world is more dangerous than sincere ignorance and conscientious stupidity.
Re:what about better SMP?
by
sQuEeDeN
·
· Score: 1
Radical? Nope... SGI makes a product like this but it is prohibitively expensive. However, the zSeries (mainframes) at IBM do this exactly. In addition, you can remove a block of processors if they fail.
--Casey.
--
Recursive(adj.): see 'Recursive'
Re:what about better SMP?
by
itzdandy
·
· Score: 1
i think this should not be a 'mainframe' technology from SGI, but something you can do with your gateway or dell.
The problem with the holistic approach
by
ahfoo
·
· Score: 1
is that it contains a troubling assumption that the whole, which he defines as the final customer solution at the software level, is known in advance and will be engineered by Big Blue.
That sounds nice at first with IBM being such a supporter of open software. But it seems to be a bit optomistic in its assumption that software development, and particularly open source software, is a homogenous and readily controlled process.
It's not just open source. I think one of the reasons Longhorn is so far off is that Bill Gates insists it will have voice recognition, but there's no way he can get that in a form that will satisfy his massive, massive audiance that is not going to learn how to speak the right way to make it work. So, MS waits hoping some little start-up somewhere will make use of the latest chip speed improvements to paste together something saleable. But in this holistic systems approach where clock speed no longer matters, that generic speech recognition system doesn't happen because those solutions aren't done by little software start-ups, they require hardware resources or as this guy refers to it holistic approaches.
Clearly the desktop is ugly at this point and the advantage is with people in markets like embedded sytems where there is still plenty of growth and a clear road map for clock speed improvements. For the desktops, it should be all about clustering and with RDMA and Gig ethernet that's not a problem. A version of RDMA called i-warp is supposed to be in typical consumer grade GigE NICs by the end of this year. And of course the GigE standard scales right up to 10GbE even with copper for short runs. So, there's another level of power to be had. The catch here is how do you appeal to the broad consumer market with this new calss of product and the answer is probably you don't, this is the end of the line for awhile.
As he mentioned, there's reconfigureable computing. But I think its a bit silly to asume something like that is closer to maturity than clustering with something like i-warp in the cards. Proprietary solutions just don't cut it. Look at Infinniband. It was backed by plenty of big names, but a few huge names isn't enough. You need a homogenous target that literally thousands of independent developers can all work towards to get real progress. Clustering does that, perhaps not as it is now, but with RDMA it will be quite different. The question is whether there is even the will to begin the process of commodifying clustered computing.
This wouldn't have anything to do with the fact that IBM has failed to be able to produce a 3 gig Power 5 chip (even though they PROMISED) because their strained silicon and silicon on insulator "tricks" failed to produce a usable 90nm process does it?
Since that announcement was recently made that 3.0 gig is currently beyond their grasp (and they tried SO hard!), this article interview seems to be nthing more than backpeddling / damage control.
IBM is putting an emphisis on alternative tactics to processor "power" now that they have clearly (and admittedly) failed in their quest for more raw speed, though you wouldn't know it from this interview. Then again, they were so SURE that their new silicon technologies would blast them past 3 gig a year or two ago, am I supposed to feel upbeat about them being able to think harder about processor design now?
-- Contrary to popular belief, coding is not all free blow-jobs and beer. Those things cost MONEY!
IBM wants Sun to open source Java... and yet (from the article):
How does this make a chip design more open? I can see where it makes it easier for people to work with IBM. But IBM still controls the processor architecture.
Cars are a good basis analogy, because they are probably the most complex systems that people have a kind of intuitive grasp of.
Perhaps even a better analogy than tach speed is to compare a Ferrari with a minivan. Which is faster over a hundred mile course? Well, if all you have to do is move a single passenger, then obviously the Ferrari. A Ferrari Enzo can cover the course in under eighteen minutes. However if you have to move seven passengers, the Ferrari has to make six and a half round trips, for a total of over 220 minutes. A minivan travelling comfortably and safely at 60 mph moves the seven passengers in a single half round tripping taking mere hundred minutes. So the minivan is more than twice as fast the Ferrari at completing this particular task.
I like this analogy for a couple of reasons. First, it introduces the idea of speed for a particular task. Secondly, IIRC, a lot of the issue with modern processors is not ther computational speed, but how quickly they can shuttle data around. This makes the passenger capacity analogy deeper than the tach analogy: you can extend it to cover why we still need mainframes, which in this analogy would be like a bus. A large bus can carry fifty to sixty passengers. A bus moving at fifty miles per hour can transport fifty passengers over our course in two hours. The Ferrari would take forty nine and one half round trips, or twenty-nine hours.
-- Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Re:Speaking of vehicles
by
Anonymous Coward
·
· Score: 0
Looks like 217.5mph, or 28 minutes on a 100 mile course.
Still, it's a really good analogy. Reminds me of the old latency vs bandwidth yarn about a 747 full of mag tapes (or 300GB HD's today).
-- My God, it's Full of Source! OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Amd has been doing this for quite a while....
by
ozzmosis
·
· Score: 1
And its true, I have a Athlon 2500+ and 2000+ and they only differ about 50MHz but the 2500+ has double the L2 cache which makes it a lot faster.
CPU: AMD Athlon(TM) XP 2500+ (1831.26-MHz 686-class CPU)
Origin = "AuthenticAMD" Id = 0x6a0 Stepping = 0
CPU: AMD Athlon(TM) XP 2000+ (1786.27-MHz 686-class CPU)
Origin = "AuthenticAMD" Id = 0x681 Stepping = 1
Just shows MHz isn't really everything, L2/L3 cache, ram speed, bus speed, have a lot to do over-all speed.
(What does Stepping = ? mean anyway??)
Re:Amd has been doing this for quite a while....
by
Antity-H
·
· Score: 1
Well, not that I know but my paranoid guess would be the following:
Ever heard of the speedstep tech ? a way for laptop CPUs to slow down, consume less power and produce less heat when they are unused... Now wouldn't that be a nice convenient way to limit the speed of a cpu so it can be sold in this or that category (in addition to cache diffs). Block the cpu s speedstep to a specific step and you are done...
of course this is a paranoid rant and can't be the real meaning of the stepping= thingy
Re:Amd has been doing this for quite a while....
by
Anonymous Coward
·
· Score: 0
Stepping refers to a rework of the silicon. You are seeing on the 2000+, AMD went back and improved the design. The 2500+ is a completely different design, but stepping 0, meaning the first release. And with the end of Socket A processors comming soon, I doubt we'll see to many steppings of the 512k Athlons.
Bare bones experience - sure - you can move resampling to an MMU
You could as you say - build a proper Image Processing Unit as a seperate chip but the economics of general purpose computers suggests that a lot will be done by the processor.
But what happens when you want a better experience - most computers for example cannot reproduce a simple slide show with a smooth fade from one image to the next in full screen resolution - this is a CPU bound problem - and very typical consumer expectation.
I think you underestimate what it takes to handle high quality images.
Images which are multi-purpose must also be of a far higher resolution than current screens - so the goal of merely showing images on a device is insuffecient.
AIK
Re:It is still relevent
by
the_2nd_coming
·
· Score: 0, Flamebait
are you a fucking retard?
what does the word CONSUMER mean to you? it means that we have a man or woman with a Sony DV camcorder and a 4 MP camera who are taking pics and videos of their kids.
Sorry - I don't mean to get into a pissing match here.
It really doesn't matter to me - but I believe there are - yes consumer - experiences which now seem the domain of WIKA digital - but like the betamax - which seemed a magic wetdream when it was first built - is not a subpar $68 dollar lost leader in grocery stores. The HDTV recorder - which seems at present to be a wetdream - must surely be considered mainstream in the very near future.
Consumers tolerate the current level - but they do want better technologies.
Look at digital cameras - 5 years ago - 2MP for $5,000 (polaroid) seemed like a fun idea - today 2MP sony is $179. And $5k will get you 11MP or more.
11MP is the equivelent of a 35mm camera - that is the resolution owned by a majority of photobuffs going in.
Why would you expect 35mm owners not to expect their digital cameras to match film?
underestimating consumer desire got the APS standard into trouble.
AIK
Power and Apple absent from spec.org?
by
3770
·
· Score: 1
I tried looking for the G5 at spec.org and IBM has a lot of entries there, and none which I could identify as the G5.
Apple didn't have any entries for 2003 or 2004 as far as I can tell.
If the G5 was the fastest on SpecInt or SpecFp I'm pretty sure Steve Jobs would have made sure they were on that list.
-- The Internet is full. Go Away!!!
The only on-topic comment on this story...
by
evilviper
·
· Score: 1
Wow... Hundreds of comments, and all their poster's own personal rant. Practically none on the actual topic.
I am just a bit confused myself...
Okay, AMD and Apple don't like the MHz rating... Sun and IBM don't like the MHz rating. And recently, Intel started developing their mobile processors into next-gen desktop processor, so they don't like the MHz rating now either.
Call this a stupid question, but doesn't that pretty much cover every company involved? So what's the hold-up, just stop including MHz ratings....
It's as if every sheep in the herd knows they are heading for the cliff, but they each keep going anyhow. It makes no sense to me.
If they all hate the MHz rating, why do they keep listing it everywhere, why do they keep advertising it, etc?
Re:The only on-topic comment on this story...
by
emorphien
·
· Score: 1
Because Mhz sells. It's hard for people to understand that one system might run at 3 ghz and its just as fast and maybe even slower than one running at lets say 2 ghz which has twice the cache or can dynamically reconfigure itself for the task at hand. That's all very confusing to the average joe, and to most CEOs.
--
Presently here, but not there.
Re:The only on-topic comment on this story...
by
evilviper
·
· Score: 1
Because Mhz sells.
It sells because they are selling it.
If they just stop listing MHz EVERYWHERE, people will begin to learn to use something else as a benchmark. Instead, every company says how stupid MHz is as a benchmark, and contiues to put their MHz ratings in big numbers on every piece of packaging and in all ads.
Re:The only on-topic comment on this story...
by
emorphien
·
· Score: 1
It's not that easy to just abandon the only thing people understand right now. If someone just stopped advertising mhz I suspect they'd meet with a rather uninterested general populace.
infinitely fast MHz (Re:Faster MHz)
by
vpetersen
·
· Score: 1
It may not help you to type faster, but I once had a computer so fast, it would execute an infinite loop in under.18 millisecs!
Your computer must have been infinitely fast then, in both, negative and positive directions of infinity. Like compiling the kernel before you hardly finished typing 'make'.
... push this clock speed marketing ploy much longer. Isn't it Amdhal's law of speedup that helps explain, going from (i'm not acutally doing the math here, just guessing) 3.0GHz to 10GHz should only give a 50% speed up or so, then 10GHz to 20GHz you get maybe 20%. I may be quite off on the acutal numbers, but you get the idea. Really what they have to figure out is how to force instructions to exectcute faster.
-- What do I have to do to get a sig around here?! www.bearscanfly.org
IBM is just lowering the level...
by
peterKslashdot
·
· Score: 1
from a bunch of chips that you can throw together to create a system, to a bunch of sections on a chip that can be thrown together to create a system-on-a-chip.
Of course, on-chip interfaces, and the whole thing, will be faster, especially when you can put best of breed components together.
As the article states, its bringing knowledge gained from high level system engineering practice down to the sub-chip level.
It's good stuff, maybe not obvious, but not mind bending either.
It's still idle a lot, when the data it needs to perform the next operation is being fetched from wherever in the memory hierarchy it happens to reside. When you quit using it for other things, the SETI program doubtless has it concentrate on running a section of code and data that all fits in your cache, so idle cycles are minimised. But don't kid yourself there are no idle cycles - there are plenty.
-- =-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
I monitor the system all the time. The CPU is solidly busy. Sure, a few instances can cause it not to be, such as a big program causing seti to swap out, or seti finishing a work unit and waiting for the next one to download. But I would not characterize these periods as "plenty" of idle cycles. Unless I am doing other things to take CPU away from seti, it's 99.9%, and when I do take cycles away, other things are using them, anyway. I think it is fair to say my system has less than 0.1% idle time.
-- now we need to go OSS in diesel cars
What I'd like to see addressed...
by
gillbates
·
· Score: 1
Is the PC's failing memory architecture.
Faster bus speeds aren't going to do it when a DMA transfer from the hard drive or ethernet card can effectively stall the processor.
What is really needed is for a memory architecture which allows the memory chip itself to move the memory, rather than having everything pass through the CPU and/or memory bus.
I've thought about this for quite some time. Consider the following memory system:
The memory chip is a 8192 x 8192 x 8 array of memory cells (64MB - put four in a machine for 256MB and 32 bit data bus). Each eight-bit byte is addressed by 8192 row and column lines.
The memory chip has three 26 bit registers: source, destination, and count. It has an additional 6 bit instruction register.
The instruction register is divided into two sections: opcode, and granularity. The opcode is a two bit field which determines one of four operations:
Read to bus
Write from bus
Internal transfer
NOP
The remaining four bits of the instruction register determine the granularity. Since their are 8192 rows, and entire row, or just part of a row of memory could be moved at a time. The contents of the six bit field contains the granularity of the operation:
0 - 1 byte
1 - 2 bytes
2 - 4 bytes
3 - 8 bytes
4 - 16 bytes
...
13 - 8192 bytes.
For regular moves, the memory works as normal: it uses the address bus to locate a byte of memory, and then places that byte on the data bus....
But, when doing large moves (rep: movsb???), the CPU figures out the length and granularity and issues instructions to the memory chip: with a 8192 byte row, the memory chip could, via a loopback circuit, move 8192 bytes per clock cycle.
With the above mentioned technique, the memory throughput for moving framebuffers and other "large" buffers on a 1 MHz (read 1000 ns RAM) chip would be 8 GB/s. With even a paltry FSB of 400MHz, that's 3.2 TB/s throughput!
Building a faster computer is not so much a matter of clock speed, but rather of efficiency. While this wouldn't improve the CPU's raw performance, it would free up the CPU to actually dedicate more clock cycles to integer and floating point workloads when performing memory intensive operations.
-- The society for a thought-free internet welcomes you.
Re:What I'd like to see addressed...
by
ajlitt
·
· Score: 1
That's great, except for the unholy number of I/Os you would need to route all across your motherboard.
The real solution is to:
1) Use DMA and an increased number of memory channels to allow the DMA controller in the system to do some of the work for you. Remember that a CPU with cache will stay off of the memory bus for significant amounts of time, letting your graphics card grab textures out of main memory, etc.
2) Make smarter peripherals. I2O is a good start, but too cost prohibitive. Modern SCSI is an excellent example, where multiple large transfers can be preprogrammed (transaction queueing) to occur when the data is ready from or ready to go to the disk. These operations then happen through DMA, requiring no CPU overhead. Ethernet is another example where coprocessing can help.
3) Use a multichannel I/O infrastructure. PCI-express is supposed to fix this. SGI had this long ago with their Avalanche bus. The idea is akin to the difference between an ethernet hub and switch. If the CPU needs to write to the ethernet buffer, and the audio needs to fetch from memory bank 1, and video needs to read from memory bank 2, then a single channel bus will slow all of these down when none of these peripherals have conflicting interests.
Re:What I'd like to see addressed...
by
gillbates
·
· Score: 1
That's great, except for the unholy number of I/Os you would need to route all across your motherboard.
Actually, the idea was to keep the transfers specific to a particular chip. That way, the 8192 byte bus would only exist as silicon pathways internal to the memory chip. The external connections would still be the familiar 64 or 128 bit data bus and 32 bit data bus.
Yes, it would mean that moving across a chip boundary would incur a substantial overhead. But considering that most systems ship with only 1 DIMM in the first place, this isn't necessarily an issue.
One could also make it such that the memory and the processor were integrated in one package - while upgradeability would be sacrificed, the performance increase would certainly justify it.
As we approach the physical limit of CPU performance, PC's are probably going to go the route of the mainframe - where upgrading means adding additional CPU's rather than replacing the existing one. In such an architecture, it's easier to tie the processor and main memory together on the same chip because it doesn't prevent a user from upgrading their memory - it just requires that they add a processor when doing so.
Incidentally, mainframe architecture already handles IO well - as each IO processor is doing its work, the main CPU's still fetch and process instructions from main memory. Writing to one area of memory doesn't preclude another processor from reading from another area - as you mentioned in your third point.
-- The society for a thought-free internet welcomes you.
Ironic, this sounds like Sun and Java
by
tkrotchko
·
· Score: 1
The arguments he makes for simultaneously calling PPC open and IBM keeping control are precisely the arguments Sun makes for not opening up Java (and are subsequently criticized by IBM).
I'm not saying its wrong, I just wanted to see if anybody else caught the irony.
-- You were mistaken. Which is odd, since memory shouldn't be a problem for you
Re:Ironic, this sounds like Sun and Java
by
Tarantolato
·
· Score: 1
The arguments he makes for simultaneously calling PPC open and IBM keeping control are precisely the arguments Sun makes for not opening up Java (and are subsequently criticized by IBM).
I suppose it's true, but then again hardware's a different kettle of fish than software. Software takes very little capital outlay to produce whereas creating even a sub-shitty microprocessor takes a big chunk of change. This means that the benefits of openness accrue much more to software, and restrictions on it are felt much closer to the bone.
Speaking of which, SPARC is an open spec as well, but that hasn't helped its adoption much.
Its a pity such a nice powerful chip like the G4 has the front side bus speed of a Pentium 2. It might have made a significant impact on the computing landscape.
Please
by
Anonymous Coward
·
· Score: 0
When it comes to transcoding video, you need speeds, and lots of it. I wish my PC (a 3.2 P4) was twice as fast. And remember, its probably about the fastest thing out there right now for this purpose.
Trucks need torque for pulling loads, racecars need horsepower for going fast.
Once the truck gets up to speed, it takes a lot less horsepower than you might think to keep it at 65.
Video editing? Compositing? 3d rendering?
by
adamgeek
·
· Score: 1
Heck, i can lock up 3+GHZ P4's to the point of every second of video taking 20+mins to render.. just with normal compositing in afx (20-30 layers, + color correction and deinterlacing).
for web surfing.. use whatever. but for any type of real video crunching, speed is everything.. most new "nice" computers probably cant even/PLAY/ 601 SD video in realtime.
Cell ???
by
Anonymous Coward
·
· Score: 0
Isn't the "Cell" processor that Sony and IBM are putting together for the next Playstation based on a PowerPC core? And isn't Microsoft using the PowerPC for the next Xbox? The Cell is supposed to be a "mult-core" processor.
There is more going on with the PowerPC family than what this article mentions. IBM has obviously come up with some nice core units that they can mix and match to build various versions of a PowerPC. Besides the normal ALU and cache units they have AltiVec units they plug in for Apple. I have read of IBM working on logic units dedicated to speeding IP communications, and ones design to provide DSP functions.
Third-parties might take standard cores and plug them together with a section of their own design to create new processors that excel at a certain task. I could see Cisco using a low-power multicore unit with special logic dedicated to IP routing and IPSEC encrypt/decrypt to create routers.
So maybe processor clock speeds won't go up as fast as in the past, but what a "CPU" can do in a clock cycle will continue to grow.
digital video.. on a mac?
by
adamgeek
·
· Score: 1
Next would be digital audio/photo/video, and that is all pretty much wrapped up by PowerMacs. You don't need to sell to those professionals and they generally have the money to fork out 4 or 5 grand for a dualie Mac every other year. Afterall, its how they earn their living!
Hmm, i do video production for a living, and I dont use a mac. Not everyone in the video world worships the apple.
Bioinformatics (processing the genome)
by
cft_128
·
· Score: 1
At my work we need the power for using programs that analyze and search the genome (blast, sim4, genewise, etc). The SQL database for storing the results is still disk bound but the algorithms are CPU bound. We normally does this on large(r) Linux servers and mini-clusters but some scientists want to do this on their desktops and even laptops (G4's being the choice, Apple and Genentech did some good work optimizing blast for the altivec).
The interview isn't reallyabout clock speed at all
by
Andy_R
·
· Score: 1
The clock speed stuff is just a tiny introductory bit, just 1 sentence long, and it really sells the article short to focus on that.
It's actually about opening up and optimising the power PC architecture - there is some fascinating stuff in there about dynamically reconfiguring the chip on the fly to optimise calculations and solve heat dissipation problems, opening up (in a very open-source like way) the chip design process to user feedback, and getting the Power chip into new markets by allowing the end user to add custom facilities to it.
-- A pizza of radius z and thickness a has a volume of pi z z a
larger cache, larger cache...
by
The+Lynxpro
·
· Score: 1
...The G5 (Power) needs a larger cache...
With AMD and Intel increasing the cache size of their processors, the IBM/Apple/Motorola triumvirate needs to increase the cache size of their processors meant for PCs, err, Macs.
Adding insult to injury is the fact that IBM manufacturers the AMD chips.
-- "Right now, somewhere in this world, Scott Baio is plowing a woman he doesn't love," - Peter Griffin, *Family Guy*
The future of computing is closed architecture--millions of cats in closed boxes. Of course the first cat computer will be enormous like ENIAC on account of the space taken up by so many cats, but eventually we will learn to shrink cats at a rate defined by Moore's Feline Law until they are so small that it is not certain whether or not they are uncertain as they should be, or merely certain. This is called the Uncertainty Recursion Problem, and is discussed here.
Just worth mentioning - the G5 CPUs in Apple's newer G5 desktops are POWER chips; the G4s were as well.
Pree Release or Article
by
Anonymous Coward
·
· Score: 0
Press release or article I couldn't tell. It just seemed like the standard pseudo-latin/techno-babble. "IBM is maintaining it position as the leading provider of current fade in cutting edge obfuscation... blah blah blah...."
Well, we're both right - what I was trying to say is a failed branch prediction means a big request to the cache; a small cache will indeed result in a lot of cache misses and so slower calls to memory.
So - a longer pipeline means you need a larger cache if you want the cache to perform comparably (have as many successful cache hits) to a CPU with a shorter pipe.
Thus: Long pipe yields larger cache. There are neutered P4s with smaller caches that very clearly demonstrate the disaster created by a small cache combined with a long pipe.
Just history repeating
by
Anonymous Coward
·
· Score: 0
Look back 15 years, and you'll find IBM taking precisely this approach to rating the comparative power of mainframes. The competition talked MIPs, IBM talked relative throughput. It was quite common for a microcode update to improve throughput and reduce MIPs at one and the same time.
Slashdot Mirror
At no point do I see him ask "If clock speed isn't important, what is?" Of course it gets answeredin a round-about kind of way.. but still.
They're only, what, almost a decade late making the observation that it's no longer as relevant for the average consumer?
Sorry, I'm a bit bitter today.
Oh, look. A story on why clock speed doesn't matter. Perhaps this is a cover-up as to why the new G5s aren't as fast as Apple promised.
SHAME on you, IBM, for causing Steve Jobs' promises not to come true.
I'm not Seth Finkelstein. I still speak the truth.
personally i like to measure the speed on how many eggs i can cook on it per minute.
my celeron can probably only do 2 or 3, i'm sure the P4 can top that though.
I mean, everybody knows its the cold cathode lights, plexiglass windows, and stickers that make it go faster.
Buy Steampunk Clothing Online!
Clock speed is good, but what I look for in a processor is that ephemeral processor attitude. Can I show it off to friends? Will my mother thinkk it's cute, or is it a little ... dangerous? I want a processor that says something about me. That I'm a rebel that won't take no for an answer. That I'm cool without trying. If a processor can't do that for me, well I'm just not interested.
Best Windows Freeware
Its interesting that apple is releasing (in july) the IBM made g5 that can go to 2.5ghz. It seems like people still care if a prosessor can go "more ghz". I think it is smart what AMD did with there 3000+ chips(or an other somthing+ chip). It makes poeple think that prosesor runs faster when it realy doesn't.
if you think clock speed is the most important measure
I think it is very important for clock speed - the crystal in my watch runs at 64k hz to keep time which is quite important. Lets see you Solaris or AMD overclockers beat that!
I do a fair amount of video editing and image manipulation, even still my two year old computer works fast enough for me...
Does the average Joe need the computing power they are given?
He doesn't really appear to offer any substantial concepts for performance improvements. Shrinking the die and upping the clock speed are the most common performance improvements because they are the most effective. Changes to the chips structure or internal coding only result in a one time 10-20% performance boost. And concepts like programmable gateways still have to follow the laws of physics.
Sure, you may be able to optimize a few very common pathways. But you simply can't optimize all of them. Thus a "perfect" algorithm for pathway adaption would again net you one of those 10-20% increases on a general processor. A dedicated machine (e.g. One attempting to calculate PI to infinity) could of course see several times the performance, but then you have to weigh an expensive programmable chip against a cheap custom chip.
Javascript + Nintendo DSi = DSiCade
... doesn't even have a clock, you insensitive clod!
is the clock speed still relevant for RISC chips too? or should that be measured differently too?
i would think the clock speed has more meaning for RISC processors.
Marge, get me your address book, 4 beers, and my conversation hat.
Sounds like the same thing AMD has been trying to convince people of for the past 5 years, while Intel has been lengthening their processor pipelines to ramp up clockspeed while effectively lowering instructions per clock. Unfortunately no one bought it when AMD was saying it, so they had to come out with their PR naming system. Let's hope that at least IBM and their significantly bigger clout can change the picture. It seems like Intel's getting on board too, it seems there are rumblings of them moving their notebook M processors to the desktop as things have gone to hell when transitioning to 90nm fabrication. (In terms of power dissipation)
in both hardware and software. While their plan doesn't seem nearly as open as GPL software, it's still a step in the right direction.
If they succeed it doesn't bode well for the x86 architecture, which seems to be a victim of it's own success. They seem to be trapped into just adding faster clocks instead of changing the architecture. They still have neat things like Centrino, but the marketing droids seem to have control over the engineers there. Every update just seems to be a faster clock speed without regards to how much it actually increases performance(I think this is evident in a lot of consumer pc's were they put in the latest and greatest pentium processor but then add in a paltry amount of RAM) I'm not saying I know more than the Intel engineers, I think they are doing a fabulous job with what they have to work with, but...I don't know where I am going with this, I'll just sit back and burn some karma now....damn ADD
to IBM's website.
I hate it when articles are posted about small, obscure companies and then I can't find their website.
Know what I like about atheists? I've yet to meet one that believes God is on their side.
I think you might be seeing a plateau on "average" users purchasing bigger, better processors or computers. If the computer they have can get the job done for them and they aren;t moving on to other computer tasks (ala video editing, rendering, etc..) they don't mind having 2+ year old computers.
Hmmm.
When I buy a PC I like to base it on figures and what it can do (AKA can it run UK 2k4, will it run HL2 for example). Rather then going "ooohhh pretty case!"
I'm funny like that, wonder why..
--- [Insert intresting Sig here]
The processor speed for marketers is comparable to the engine size wars in the 60s/70s. If I say I have a 402 (6.6L) in my Chevelle and Bob next door has a (snicker) 350 (5.7L) in his Nova, my car gets the approving nods, but may not be faster since the Nova is lighter. Now compare said Chevelle w/today's Z06 'vette. Little 'wimpy' vette has just a 5.7L, but kicks the snot outta the Chevelle in performance. IBM, and other marketing 'geniuses', need to name their products to entice the 'mine is bigger' crowd. Right now, in the consumer computer realm, GHz talks. Most non-IT people I know will spout the "My PC is 4GHz - what's yours?" mantra when a 2.8 Opteron w/SCSI320 will kick its butt. The enlightened will know, but 'tis the general ignorant masses that have the buying power.
...telling you that it's the GIRTH that counts!
Best Buy can have you arrested
Corporate datacenters are now filling up with half-full racks because cooling & power requirements are through the roof. You end up being unable to increase compute resources because you have to put in fewer of the faster systems, gaining you nothing.
So hey, if you're listening harware vendors, see if you can't simply make the dang things run cooler on less power before you speed them up!
Apple has been trying to get this message out for a while. We had a story a few months ago (lazy me, no link for you) about how Intel was dropping the clock from the branding of the processors.
Clock speed really does not have a direct correlation to computer speed anymore. It seems like we will see more of the trend of newer, better technology that runs at a lower rate but executes a lot more in one tick, so it is much faster. It seems that it will start at 1GHz and move up to 3 before somebody gets a new idea, makes a new "slow" processor and starts it over...
Even on the base CPU, the most important metric, I find, is "MIPS per watt". That's what determines how much horsepower you can get off a given amount of cooling, which is the real limiting factor for CPU speed.
Mirroring ibm.com should probably be modded Redundant :D
So I "dumb" my speach down a bit and give it again. The masses don't want to know how processors work, they don't want to know about architecture, they want an even base line to measure performance. Most people think the Comp Usa rep is ripping them off and they are trying to feel good about their purchase, being an un-educated consumer.
By buying the high clock speed they can compare it to their neighboors and in their heads they have a Super-Fast PC.
I'd like to note that most people I talk to look at AMD like most people look at a Yugo. (remember those cars?) In spite of my advice that an AMD is like a new Honda for $2,000.
That's my 2 cents
I boycott signatures
There are many applications, like software defined radios and televisions, that require huge amounts of number crunching.
Mea navis aericumbens anguillis abundat
The processor shouldn't just lock up like that. Perhaps the heavy load is overheating it?
Clock speed has never been the definitive CPU performance measurement. As everyone knows it is the Bogo MIP.
-- Thou hast strayed far from the path of the Avatar.
Good question.
I contribute my spare computing cycles to distributed computing efforts.
The distributed.net client, at least, does not require the computer to be switched on 24/7 either.
Even playing DVDs only take up about 25% of the CPU time, and I've had no problems with overheating. I sometimes think about all that A/C power and computer cycles being wasted at the university computer rooms.
Maybe O.S. vendors could include a voluntary option doing install to contribute your computing power. They could then re-sell that computing to users who need to do heavy calculations that are not time-critical or secret.
P.S. IBM is hot on grid computing, but they don't all seem to believe the processing power also can be contributed by individuals.
Irene KHAAAAAAN!
says processor chip speed is old news Half the posts on /. already say this... I saw two different posts today from different topic headers, one was from the Apple post on the G5
I mod down so you can mod up. Your welcome.
Converting a bunch of AVI to VOB format so I can burn them takes like 9 hours for 80 minutes of resulting video on my 1GHz P3. A 1.6Ghz P4(it's in a laptop w/ speedstep...not sure of name) does it in 3.5 hours.
Blar.
Didn't AMD take this approach some years back? They have to name their processors to sounds like pentium clock speed ratings, but they have been preaching the idea that clock-speed is not the sole issue for years. I know IBM is technical leader, but it just smells like IBM, like Intel, are jumping on the AMD bandwagon, but they aren't calling it the AMD bandwagon.
Great ideas often receive violent opposition from mediocre minds. - Albert Einstein
that its not the clock speed that matters, but how high the number is beside the processor that matters....duh.
would want there clock to go at more than (1/60)Hz? Im trying to get MORE SLEEP, not less. Geeze, just increasing clock speed to 1 Hz, means that in a typically day, I would get .1333 hours of sleep on a good day.
...you're getting a Dell.
(Sorry, slow day at work)
Moderation Total: -1 Troll, +3 Goat
"...because if you let the smoke out, it stops working."
- my friend, when he's asked what's inside a transistor.
SIGERR: laziness exceeds quota
business is often about defining your strategy for approaching new business. ibm is stating that openess will benefit their business. others have recognized this holistic approach to systems design, which from the beginning of computing was really required for high end/specialized systems. some wanted openess for additional reasons, such as freedom in beer and speech. so the same stick has simply been picked up to use in a more competetive environment against businesses not capable of integrating both sides of the computing environment.
so long as everybody has their needs met, it's a good thing(tm).
I don't care if they want to name it by processing speed, bus speed, or hell how much donkey speed it has - i just want it to be consistent!!! That means give me Intel processor 1, then the future processor named Intel Processor 2 WILL be better then the first one, and the third one will be better, etc. Stop coming out with funky new names to confuse me (and the less informed computer users, which i am probably in that category). If I want to look at the specs I will, but at least make it easy for people to realize which processor is the latest and greatest! I think the worst case for this is some of the graphics cards - whats better the 9600 version or the FX version or the super version or the crack version? Come on, have some sympathy on us ----damn marketers!
I mod down so you can mod up. Your welcome.
In other words, the computer is helping to design computers that are faster than itself?
But then you go out to design an even better computer? That can then design an even better computer?
How ironic it would have been if the only application of computer technology was creating new computers.
"I speak of none but the computer that is to come after me, A computer whose merest operational parameters I am not worthy to calculate. And yet I will design it for you."
Irene KHAAAAAAN!
The old MHz measurement was nice for a time, but it just wasn't measuring what computer performance geeks find important. Just like car performance enthusiasts like to talk about the horsepower ratings of their engines, what is really important is a measurement of how the system really performs. That's why a quarter-mile timing can be more informative about a car's performance than just looking at the car's horsepower and torque rating.
So, the new processor measurement that gets right to the heart of what's important for a CPU performance geek is going to be henceforth the PU. The PU stands for 'penis unit' and it indicates to their fellow cpu performance geeks how big their dick is relative to everyone else's.
Fascism trolls keeping me up every night. When I starts a preachin', he HITS ME WITH HIS REICH!
The POWER architecture isnt really for the average Joe's computer, it is for servers. In servers, many tasks are done by coprocessors and independant subsystems without taxing the CPU. The extra CPU performance is now suddenly needed because IBM keeps encouraging ISV'S to write for Websphere, in Java, so you now need 10 times more memory and CPU performance than you previously did to perform the same task. In servers, the worst bottleneck at the moment are afaik still the moving parts in the disks and tapes.
The PPC is a cousin of the embedded version of the chip, where the performance per watt power usage is relevant. It is hugely successful.
Sales of Apples with desktop POWER chips aren't really significant. Although IBM aren't ready to admit it yet, the battle for the desktop is long over. No amount of performance advantage is going to outweigh the main advantage of the x86 architecture there: backward compatibility, preserving the value of past investments in software for the customer. IBM should know this, as they still make their zSeries mainframes compatible with the 40+ year old 360 architecture for the same reason.
In the PC, unlike most servers, most everything goes through the CPU, which is why for the average Joe raw CPU performance _does_ matter.
In which case you're more interested in the disk and memory subsystems than in raw CPU speed. From what I've heard, VMWare is also just as much a memory hog as CPU hog, too.
For both of these cases, seems to me that your incremental dollar would be better spent on more memory and/or faster FSB instead of a simply-higher clock speed CPU.
Obvious caveats:
Once you hold your working set in RAM, there may be little point in more memory.
Assuming you're 32bit, the 4G+4G patch may hurt more than it helps, and you may be max'ed in the 1-4G range, but how big is your source tree?
Higher FSB and clock speed are somewhat related, but there's still wiggle room on both Intel and AMD.
The living have better things to do than to continue hating the dead.
Addressing a fundamental shift in the landscape of technology, a significant shift in the trajectory of basic technology. The rate of performance enhancement is becoming impacted as holistic design. You have to have a means by which you proactively and holistically address that extraordinary event. I am hoping that people really understand the sort of discontinuity we are talking about the capability for the chip to physically morph, one can easily envision the system autonomically issuing a command moving off to an entirely different plane.
The world is full of incredibly bright people.
What other applications besides games really tax the CPU right now?
Are you planning on a career in the computing field? If so, please send me your name and address info so I'll know never to hire you.
If you think the only thing a fast CPU is used for these days is playing "Unreal Tournament", you have no business managing (or even logging in to) the kinds of UNIX boxes some of us deal with on a daily basis. Yes, a faster CPU (or 24 of them) is more than necessary for some of the applications we run in the Real World.
No, we're not beta-testing the next "Doom". We're building aircraft.
- A.P.
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
AMD and even Cryix before them have been telling us this for years..
You have 5 Moderator Points!
Which Helpless Linux zealot/MS basher do you want to mod down today?
IBM is attempting to change the rules of performance measurement. They are doing this by educating their customers. Inherently, people want a single performance metric that says X is better than Y because this perfromance metric says so.
IBM would prefer customers to come to them and ask IBM, "Which processor is better?" rather than rely on an external, easily verifiable, though not accurate, single number indicator.
The truth, as we all know, is that there is no single metric since each processor has strengths and weaknesses and various applications rely on these strengths differently.
They are also opening their processors to the end users a little more, almost as a jab at Intel. Intel has microcode, but you'll never see it or get to modify it. But the very presence of microcode in almost every modern general purpose CPU means that performance can be enhanced and tailored for each application with very little processor change.
So IBM is letting people get closer with the processor to enhance performance with very little risk or effort.
The kicker is that it's not simple, so only a few large manufacturers and some dedicated homebrewers will really have anything to show for it.
Thus it's a marketting ploy intended to raise questions about current performance metrics in the minds of indecisive consumers.
But then, when has the CPU war ever been about anything but marketting?
-Adam
This is very significant, and not because of the clock cycle bit. 90% of the comments are responding to the article title, not content. Typical.
The open architecture ecosystem can have a far reaching fundamental effect on potential market-space of CPU dependent products.
This is far bigger news than any of Sun's open sourcing or Brown v. Tanenbaum.
One thing I've learnt over the years is that just because a company talks about open standards doesn't mean that there is.
If I can build my own fab and start producing exact copies of IBM's design and sell them without even seeking permission, then I'll believe it. (same with the sparc) Until then, it's all talk.
This is all marketing. Where does having the fastest processor really matter? Gamers, maybe, but what gamers buy a Mac with an IBM chip to play games?
Next would be digital audio/photo/video, and that is all pretty much wrapped up by PowerMacs. You don't need to sell to those professionals and they generally have the money to fork out 4 or 5 grand for a dualie Mac every other year. Afterall, its how they earn their living!
99% of the people surf the web, check e-mail, pay their bills online, write letters and occassionally play a game more complicated than free cell. You can do all that with the slow-assed 733 mhz G4 I have (with no L3 cache to boot)!
These people must also feel the need to drive a Ferrari to the convenience store.
"...a fundamental shift in the landscape of technology..."
ZZZZZzzzzzzz.....
There is a trade-off between speed, reliability, cost, and re-programmability.
SRAM types
Are re-programmable but require a rather slow serial load at boot-up. Reliability in embedded systems leaves something to be deisired since any brownout-induced glitch can create errors that are even worse (harder to recover from) than software glitches because wired logic doesn't have anything equivalent to code checksums or interrupt vectors. Well-paid FPGA designers are versed in the arcane art of self-verifying logic.
EEPROM types
Come alive at boot up and are much more resistant to glitches. Their performance, however, is slow. And you have limited (100,000 maybe) rewrite cycles.
Anti-fuse types
are made by Actel. They have the highest performance and best density. They come alive at boot up and are dead-nuts reliable under the worst of conditions; for example, properly qualified, they can survive the cosmic radiation in spacecraft that would leave other types toasted. The big drawback: the anti-fuse process, which works by melting diodes into short-circuits, is not eraseable.
Desktop systems (say, an add-on FPGA card) would be best served by SRAM types, since you already have a processor that requires gluttinous gobs of puritanically clean DC power. Basement hardware hackers would be better served by EEPROM or anitfuse types (depending on performance requirements), since they don't require super-expensive exotic design software.
But it is a big deal. First of all, a lengthened pipeline will result in more costly branches. As a result, more transistors will be required to beef up the branch prediction + cache. And all these extra transistors will add to the power consumption and manufacturing cost of the chip. So you end up with a chip that uses far more transistors then an equivalent chip with a smarter instruction set / design (just compare transistor counts between the P4ee, Power, and Athlon64 of the same speed - the extra long pipeline of the P4ee hurts).
Most of the statements you made were dead on and I would generally consider your post to be a good one. But I can't help but imagine what todays CPUs would be like if everyone used an open source OS that could be run on any CPU architecture. This would result in a more competitive market by making it easier for new companies to enter the market with cool new designs. I'm willing to bet that x86 chips would have died off long ago.
FPGAs are useful, but very few applications reprogram them while in use. I doubt that's what they intend.
Machines have been built into which you could load application-specific microcode. The DEC VAX 11/780 had that feature, although nobody ever used it. Shader programs in GPUs are probably the most successful example of this idea.
They'll probably export some internal tuning parameters, like the instruction/data ratio for the cache or the depth of instruction prefetch, produce some benchmarks showing that if you tune those performance goes up 20% or so, and hype that into "reconfigurability".
here is a thought.
When you buy a car, you don't just consider how fast it goes, you consider fuel economy/comfort/quality etc. this could be applied to CPUs
for instance, you buy a new DeLL/HP/whatever, the machine has 3 numbes on it. 12/16/65 - which means, 12 is the general office app benchmark, 16 is the gaming benchmark, and 65 is the mean power usage of the machine.
so a good office machine is a 14/2/30, but if you are playing games, you need a 6/26/130, you don't care as much about the power bill or how fast office computes a6:c6*c14 whatever. these numbers would be linear, as in the nex-gen would just have higher numbers.
Each computers label would ave a description of the rating above the label saying "look at the killer gamer system" or whatever.
I can see the arguement of the system being confusing, but i'd take the least confusing method that was effective, and i think this would be effective.
------------------
Something like this could translate over to server side with web/fileserving/powerreq or something, but it would allow companies like AMD and IBM who have not pushed the MHz myth to the extreme to allow their product to compete on merrit not Mhz.
thoughts??
will they help me to type faster?
...in your post. I would never have found it otherwise!
the whole landscape of computing could shift to more SMP capable systems. Ideally, you should have some interconnect for CPU's that was more modular. Have a high speed interconnect for just processors OR have a universal high speed interconnect for everything including processors. have the system see a CPU simply as somthing that run processes, and have some small memory on the chip that let the system know what that CPU could process and how man MIPS/FLOPS/whatever it was capable of running.
You would boot up your OS and have X amount of processing power, then you could plug in a CPU card into a spare slot and it would hotplug and be added into the processing pool for whatever instructions it was capable of processing.
ideally, you could run multiple intruction sets on any given system, because when you plugged in a new processor, it would allow those new instructions to be executed, your powerpc programs would no longer segfault OR they would not have to run in emulation space.
you could also have the system use power management to shut off extra processors and put their abilities into a "powerup if needed" state, so you would only power up for floating point processor if needed, and only run it at the Mhz needed for the task at hand.
its radical, but it could be a good idea.
thoughts?
is that it contains a troubling assumption that the whole, which he defines as the final customer solution at the software level, is known in advance and will be engineered by Big Blue.
That sounds nice at first with IBM being such a supporter of open software. But it seems to be a bit optomistic in its assumption that software development, and particularly open source software, is a homogenous and readily controlled process.
It's not just open source. I think one of the reasons Longhorn is so far off is that Bill Gates insists it will have voice recognition, but there's no way he can get that in a form that will satisfy his massive, massive audiance that is not going to learn how to speak the right way to make it work. So, MS waits hoping some little start-up somewhere will make use of the latest chip speed improvements to paste together something saleable. But in this holistic systems approach where clock speed no longer matters, that generic speech recognition system doesn't happen because those solutions aren't done by little software start-ups, they require hardware resources or as this guy refers to it holistic approaches.
Clearly the desktop is ugly at this point and the advantage is with people in markets like embedded sytems where there is still plenty of growth and a clear road map for clock speed improvements. For the desktops, it should be all about clustering and with RDMA and Gig ethernet that's not a problem. A version of RDMA called i-warp is supposed to be in typical consumer grade GigE NICs by the end of this year. And of course the GigE standard scales right up to 10GbE even with copper for short runs. So, there's another level of power to be had. The catch here is how do you appeal to the broad consumer market with this new calss of product and the answer is probably you don't, this is the end of the line for awhile.
As he mentioned, there's reconfigureable computing. But I think its a bit silly to asume something like that is closer to maturity than clustering with something like i-warp in the cards. Proprietary solutions just don't cut it. Look at Infinniband. It was backed by plenty of big names, but a few huge names isn't enough. You need a homogenous target that literally thousands of independent developers can all work towards to get real progress. Clustering does that, perhaps not as it is now, but with RDMA it will be quite different. The question is whether there is even the will to begin the process of commodifying clustered computing.
This wouldn't have anything to do with the fact that IBM has failed to be able to produce a 3 gig Power 5 chip (even though they PROMISED) because their strained silicon and silicon on insulator "tricks" failed to produce a usable 90nm process does it?
Since that announcement was recently made that 3.0 gig is currently beyond their grasp (and they tried SO hard!), this article interview seems to be nthing more than backpeddling / damage control.
IBM is putting an emphisis on alternative tactics to processor "power" now that they have clearly (and admittedly) failed in their quest for more raw speed, though you wouldn't know it from this interview. Then again, they were so SURE that their new silicon technologies would blast them past 3 gig a year or two ago, am I supposed to feel upbeat about them being able to think harder about processor design now?
Contrary to popular belief, coding is not all free blow-jobs and beer. Those things cost MONEY!
Cars are a good basis analogy, because they are probably the most complex systems that people have a kind of intuitive grasp of.
Perhaps even a better analogy than tach speed is to compare a Ferrari with a minivan. Which is faster over a hundred mile course? Well, if all you have to do is move a single passenger, then obviously the Ferrari. A Ferrari Enzo can cover the course in under eighteen minutes. However if you have to move seven passengers, the Ferrari has to make six and a half round trips, for a total of over 220 minutes. A minivan travelling comfortably and safely at 60 mph moves the seven passengers in a single half round tripping taking mere hundred minutes. So the minivan is more than twice as fast the Ferrari at completing this particular task.
I like this analogy for a couple of reasons. First, it introduces the idea of speed for a particular task. Secondly, IIRC, a lot of the issue with modern processors is not ther computational speed, but how quickly they can shuttle data around. This makes the passenger capacity analogy deeper than the tach analogy: you can extend it to cover why we still need mainframes, which in this analogy would be like a bus. A large bus can carry fifty to sixty passengers. A bus moving at fifty miles per hour can transport fifty passengers over our course in two hours. The Ferrari would take forty nine and one half round trips, or twenty-nine hours.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
And its true, I have a Athlon 2500+ and 2000+ and they only differ about 50MHz but the 2500+ has double the L2 cache which makes it a lot faster.
CPU: AMD Athlon(TM) XP 2500+ (1831.26-MHz 686-class CPU)
Origin = "AuthenticAMD" Id = 0x6a0 Stepping = 0
CPU: AMD Athlon(TM) XP 2000+ (1786.27-MHz 686-class CPU)
Origin = "AuthenticAMD" Id = 0x681 Stepping = 1
Just shows MHz isn't really everything, L2/L3 cache, ram speed, bus speed, have a lot to do over-all speed.
(What does Stepping = ? mean anyway??)
I can't believe you know what a DSP is!!!
This is what Apple should have put in their G5 PowerMac's instead of a "crippled" PowerPC.
The hottest consumer electronic product right now may well be the digital camera.
Given that it takes a lot more cycles to process an image - than say - a spreadsheet - I think the consumer very much wants processing speed.
The next hot item will be likely be digital videos actually going somewhere - other than the shoebox
This will be even more cycle intensive.
The market responds to technology in fits and starts - but the analysis says many consumer products are still throttled by the speed on consumer PC's
People buying HDTV today will want HDTV camcorders tomorrow.
A computer that can render HDTV signals from disk will be a benchmark standard in only a few years.
The idea that spreadsheet surfers are satisfied with clock speed is quiet impressive - and passe.
AIK
I tried looking for the G5 at spec.org and IBM has a lot of entries there, and none which I could identify as the G5.
Apple didn't have any entries for 2003 or 2004 as far as I can tell.
If the G5 was the fastest on SpecInt or SpecFp I'm pretty sure Steve Jobs would have made sure they were on that list.
The Internet is full. Go Away!!!
Wow... Hundreds of comments, and all their poster's own personal rant. Practically none on the actual topic.
I am just a bit confused myself...
Okay, AMD and Apple don't like the MHz rating... Sun and IBM don't like the MHz rating. And recently, Intel started developing their mobile processors into next-gen desktop processor, so they don't like the MHz rating now either.
Call this a stupid question, but doesn't that pretty much cover every company involved? So what's the hold-up, just stop including MHz ratings....
It's as if every sheep in the herd knows they are heading for the cliff, but they each keep going anyhow. It makes no sense to me.
If they all hate the MHz rating, why do they keep listing it everywhere, why do they keep advertising it, etc?
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
What the heck does that mean? IBM will release chips that will optimally name themselves, automatically?
It may not help you to type faster, but I once had a computer so fast, it would execute an infinite loop in under .18 millisecs!
Your computer must have been infinitely fast then, in both, negative and positive directions of infinity. Like compiling the kernel before you hardly finished typing 'make'.
... push this clock speed marketing ploy much longer. Isn't it Amdhal's law of speedup that helps explain, going from (i'm not acutally doing the math here, just guessing) 3.0GHz to 10GHz should only give a 50% speed up or so, then 10GHz to 20GHz you get maybe 20%. I may be quite off on the acutal numbers, but you get the idea. Really what they have to figure out is how to force instructions to exectcute faster.
What do I have to do to get a sig around here?! www.bearscanfly.org
from a bunch of chips that you can throw together to create a system, to a bunch of sections on a chip that can be thrown together to create a system-on-a-chip. Of course, on-chip interfaces, and the whole thing, will be faster, especially when you can put best of breed components together. As the article states, its bringing knowledge gained from high level system engineering practice down to the sub-chip level. It's good stuff, maybe not obvious, but not mind bending either.
... except when it's connecting to the SETI server to exchange another work unit.
now we need to go OSS in diesel cars
It's still idle a lot, when the data it needs to perform the next operation is being fetched from wherever in the memory hierarchy it happens to reside. When you quit using it for other things, the SETI program doubtless has it concentrate on running a section of code and data that all fits in your cache, so idle cycles are minimised. But don't kid yourself there are no idle cycles - there are plenty.
=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Friends don't let friends enable ecmascript.
Is the PC's failing memory architecture.
Faster bus speeds aren't going to do it when a DMA transfer from the hard drive or ethernet card can effectively stall the processor.
What is really needed is for a memory architecture which allows the memory chip itself to move the memory, rather than having everything pass through the CPU and/or memory bus.
I've thought about this for quite some time. Consider the following memory system:
Building a faster computer is not so much a matter of clock speed, but rather of efficiency. While this wouldn't improve the CPU's raw performance, it would free up the CPU to actually dedicate more clock cycles to integer and floating point workloads when performing memory intensive operations.
The society for a thought-free internet welcomes you.
The arguments he makes for simultaneously calling PPC open and IBM keeping control are precisely the arguments Sun makes for not opening up Java (and are subsequently criticized by IBM).
I'm not saying its wrong, I just wanted to see if anybody else caught the irony.
You were mistaken. Which is odd, since memory shouldn't be a problem for you
Its a pity such a nice powerful chip like the G4 has the front side bus speed of a Pentium 2. It might have made a significant impact on the computing landscape.
When it comes to transcoding video, you need speeds, and lots of it. I wish my PC (a 3.2 P4) was twice as fast. And remember, its probably about the fastest thing out there right now for this purpose.
Trucks need torque for pulling loads, racecars need horsepower for going fast.
Once the truck gets up to speed, it takes a lot less horsepower than you might think to keep it at 65.
Heck, i can lock up 3+GHZ P4's to the point of every second of video taking 20+mins to render.. just with normal compositing in afx (20-30 layers, + color correction and deinterlacing).
/PLAY/ 601 SD video in realtime.
for web surfing.. use whatever. but for any type of real video crunching, speed is everything.. most new "nice" computers probably cant even
Isn't the "Cell" processor that Sony and IBM are putting together for the next Playstation based on a PowerPC core? And isn't Microsoft using the PowerPC for the next Xbox? The Cell is supposed to be a "mult-core" processor.
There is more going on with the PowerPC family than what this article mentions. IBM has obviously come up with some nice core units that they can mix and match to build various versions of a PowerPC. Besides the normal ALU and cache units they have AltiVec units they plug in for Apple. I have read of IBM working on logic units dedicated to speeding IP communications, and ones design to provide DSP functions.
Third-parties might take standard cores and plug them together with a section of their own design to create new processors that excel at a certain task. I could see Cisco using a low-power multicore unit with special logic dedicated to IP routing and IPSEC encrypt/decrypt to create routers.
So maybe processor clock speeds won't go up as fast as in the past, but what a "CPU" can do in a clock cycle will continue to grow.
Next would be digital audio/photo/video, and that is all pretty much wrapped up by PowerMacs. You don't need to sell to those professionals and they generally have the money to fork out 4 or 5 grand for a dualie Mac every other year. Afterall, its how they earn their living!
Hmm, i do video production for a living, and I dont use a mac. Not everyone in the video world worships the apple.
At my work we need the power for using programs that analyze and search the genome (blast, sim4, genewise, etc). The SQL database for storing the results is still disk bound but the algorithms are CPU bound. We normally does this on large(r) Linux servers and mini-clusters but some scientists want to do this on their desktops and even laptops (G4's being the choice, Apple and Genentech did some good work optimizing blast for the altivec).
Underloved Movies and Pub Quiz: donotquestionme.org
The clock speed stuff is just a tiny introductory bit, just 1 sentence long, and it really sells the article short to focus on that.
It's actually about opening up and optimising the power PC architecture - there is some fascinating stuff in there about dynamically reconfiguring the chip on the fly to optimise calculations and solve heat dissipation problems, opening up (in a very open-source like way) the chip design process to user feedback, and getting the Power chip into new markets by allowing the end user to add custom facilities to it.
A pizza of radius z and thickness a has a volume of pi z z a
With AMD and Intel increasing the cache size of their processors, the IBM/Apple/Motorola triumvirate needs to increase the cache size of their processors meant for PCs, err, Macs.
Adding insult to injury is the fact that IBM manufacturers the AMD chips.
"Right now, somewhere in this world, Scott Baio is plowing a woman he doesn't love," - Peter Griffin, *Family Guy*
The future of computing is closed architecture--millions of cats in closed boxes. Of course the first cat computer will be enormous like ENIAC on account of the space taken up by so many cats, but eventually we will learn to shrink cats at a rate defined by Moore's Feline Law until they are so small that it is not certain whether or not they are uncertain as they should be, or merely certain. This is called the Uncertainty Recursion Problem, and is discussed here.
English is easier said than done.
Just worth mentioning - the G5 CPUs in Apple's newer G5 desktops are POWER chips; the G4s were as well.
Press release or article I couldn't tell. It just seemed like the standard pseudo-latin/techno-babble. "IBM is maintaining it position as the leading provider of current fade in cutting edge obfuscation... blah blah blah... ."
and less need for a large cache (big pipeline means frequent cache hits)
Don't you mean that a big pipeline means a bigger penalty for cache misses? Pipeline length has no effect on the cache hit ratio.
± 29 dB
Look back 15 years, and you'll find IBM taking precisely this approach to rating the comparative power of mainframes. The competition talked MIPs, IBM talked relative throughput. It was quite common for a microcode update to improve throughput and reduce MIPs at one and the same time.