Fujitsu's supercomputer-ready Venus CPU said to be "world's fastest"
Due to the intrinsic limitations of machine translation, it's hard to say exactly what makes Fujitsu's latest supercomputer the "world's fastest," but we'll hesitantly believe for the time being. We're told that the SPARC64 VIIIfx (codename Venus) can churn through 128 billion calculations per second, which supposedly bests the current champ -- a chip from Intel -- by 2.5 times. An AP report on the matter states that Fujitsu shrunk the size of each central circuit, which in turn doubled the number of circuits per chip. 'Course, this beast won't be ready for supercomputer work for several years yet, giving the chip maker's biggest rivals plenty of time to sabotage its moment in the limelight.
[Via Physorg]
[Via Physorg]























DO WANT.
it plays doom
IT IS DOOM!!!
So now we're all DOOMed?
Can it play Doom? Is that the best you guys can do?
I can't wait till they release the first home quantum computer in about 20 years so someone can ask that and I can reply...
...it can play all of the Dooms, Crysises and Quakes simultaneously
it's about the size of the chip that's in Arnold's head in Terminator 1 and 2...
So yes, it could very well be Doom!
Wrong. It's SPARC, that means it doesn't use the x86 architecture. So without a recompile, it will not play dom.
quantum physics, why would anyone need a quantum desktop...
The only functional QCs have the power of a calculator and need to be kept at 12 kelvin...
@Penguin
Actually man, and its not usually my thing to split hairs, but since you are...
The doom source was released in... what was it... 97? And has been mad pretty portable over the years, so yes it does run on sparc. But more importantly, if that chip is as fast as they say it is, it can do better then just run the x86 version of Doom (via an x86 emulator) it could probably do some very nice after effects and upscale it.
... sorry to do that.
It munches through Crysis like a fat lady at an all-you-can-eat buffet
I'm happy you beat the guy below you who had the stock question.
Cheers!
Actually, it won't run Crysis at all because it uses a custom instruction set rather than standard x86 or x64 and Crysis won't run in any current commercial VM.
@DWells55:
I was providing levity while preempting the inevitable comments "can it run XYZ?" :) I would think most people with a remote interest in processor architectures know Crysis won't run on this unless Crytek recompiles it (yes, I'm vastly understating the matter, while I'm sure they aren't using any x86/x64 asm (lending hope to portability), Crysis is inextricably tied to the systemic cancer we call DirectX)
Yeah, I kind of figured that, but felt like taking it literally and being a buzzkill.
Too bad OpenGL doesn't do more to compete with DirectX. Stupid proprietary Microsoft APIs. If it weren't for DirectX, we'd be seeing a lot more games for Mac and Linux.
I wonder how a emulator/dynarec would work on the venus ......... considering the SIMD it might even be able to handle GPU emulation or make software rendering crazy fast or somethign like that
But will it play Crysis?
Stop it already! Crysis runs pretty decent on almost anything, and that joke is two years old!
just down rate them
not smart. i think getting low ranked gives them their life energy. just find where they live and take their flesh light and blow up doll away
i voted up 'cos it was funny.
oh
Bleh what good will it do in a supercomputer? Install it in my PC instead! I'll run some Folded@home proggie for CPU and maybe it can finally match the speed of GTX 280 xD
Go Fujitsu!
Go AMD!
What about the Cell processor? it is said that it has 256 Gflops.
The instruction set etc. matters. For example the high end AMD and nVidia GPU chips can hit 1,000 GFLOPS peak .. whereas this one is at around 128 GFLOPS peak.
I wouldn't run around saying they have the world's fastest CPU yet without benchmarks etc.
I would love to play solitaire on that thing
id love to see this baby run notepad.exe phwaaaaaar
The heart of Skynet everybody..
Would anyone bet against Intel increasing its own chip performance more than 2.5 times "within several years"...
Several years... 3+? Moore's law is ticking away. They best hope the calculation power isn't close to the final product performance.
That huge chip from before doesn't beat this?!
http://www.engadget.com/2009/05/15/huge-new-intel-processor-revealed/
This might be 2.5 times faster than Intel's chip but it's no threat to them. Intel will just pay computer makers not to use it.
It'll be used in supercomputing, where Intel doesn't have much pull.
As for desktops/workstations, AMD already took care of SPARC in that market - note that Sun rather quickly ditched SPARC workstations once the Opteron came out.
Imagine....I could pwn every single Noob in all the servers for World of Warcraft....
I'll be the new Chuck Norris with this baby!
You cannot compare this highly parallel SPARC to a quad-core Nehalem. They are each the "fastest" in their respective usage scenario.
Doesn't this kind of go counter to the current trend in CPU design for super computers?
I mean it's amazing what they have been able to do, but that is one huge chip, which means yield per wafer is not gonna be all that high, which means it is going to be an extremely expensive chip.
But that's not even the problem that I see... Super computers aren't a ONE CPU system - I would think that it is still better to gang a bunch of cheaper chips together (even if individually they are slower), than having less (but more powerful) chips.
This is 2.5x the speed of the fastest comparable Intel chip, but if the Intel chip is 1/3 the price and just as scalable, then Intel would still make the better chip to use.
Actually in reality it would run today's games like crap. that's not what its designed for.....