Intel Larrabee finally hits 1TFLOPS - 2.7x faster than nVidia GT200!

[Olivon]

whats is impresive in this ati already reached teraflop i think 1.8 or something

It's concerning SGEMM 4K x 4K performance

1. Intel Larrabee [LRB OC, 45nm] - 1006 GFLOPS
2. EVGA GeForce GTX 285 FTW - 425 GFLOPS
3. nVidia Tesla C1060 [GT200, 65nm] - 370 GFLOPS
4. AMD FireStream 9270 [RV770, 55nm] - 300 GFLOPS
5. IBM PowerXCell 8i [Cell, 65nm] - 164 GFLOPS
6. i7 XE 975 - 101 GFLOPS

[eric66]

It's concerning SGEMM 4K x 4K performances

there is post there saying this 'Anyways I've read that SGEMM on AMD/ATI 5870 has reached 67 % of theoretical performance.

This would mean that the 5870 is doing 0.67*2.72 TFLOPs ~= 1.8 TFLOPs on SGEMM. Almost twice the peformance of LRB.'

don't know if its true or note though

[Russian]

Thx for the video Face

ajaidev : corrected

The video was linked in the article.

[Aerou]

This will be mighty WCG cruncher

That depends, but I like the thinking

[haylui]

AMD is going to have new GPU next year....lolz...

[JumpingJack]

there is post there saying this 'Anyways I've read that SGEMM on AMD/ATI 5870 has reached 67 % of theoretical performance.

This would mean that the 5870 is doing 0.67*2.72 TFLOPs ~= 1.8 TFLOPs on SGEMM. Almost twice the peformance of LRB.'

don't know if its true or note though

This is a good point, if Larrabee had been productized early this year say Q1, it would have been a formidable competitor (under the assumption that this is the true throughput performance), but it has already been exceeded, so it may end up only playing in the midrage.

[dinos22]

i love how they all got so excited hitting the Teraflop hahaha

[Serpentarius]

This is a good point, if Larrabee had been productized early this year say Q1, it would have been a formidable competitor (under the assumption that this is the true throughput performance), but it has already been exceeded, so it may end up only playing in the midrage.

well .. on the economic side ... midrange gpu is a lucrative bite
by just reducing both ATi and Nvidia sales by 20% each ... it can cause quite a serious damage to them. most users purchase midrange gpus u know

well .. imho it could cause both ATi and Nvidia to end up selling highend gpus ...
but since we all know that intel has the tendency to "encourage" mobo manufacturers to build what they want (or else)

most probrably we can see a great shift in this?
the flaming gpu wars ... lol

[Smartidiot89]

the flaming gpu wars ... lol

We'll have 3 blocks of fanboys arguing that their card is better

[JumpingJack]

well .. on the economic side ... midrange gpu is a lucrative bite
by just reducing both ATi and Nvidia sales by 20% each ... it can cause quite a serious damage to them. most users purchase midrange gpus u know

well .. imho it could cause both ATi and Nvidia to end up selling highend gpus ...
but since we all know that intel has the tendency to "encourage" mobo manufacturers to build what they want (or else)

most probrably we can see a great shift in this?
the flaming gpu wars ... lol

Personally, I think a 3rd major play in GPU is a good idea, I hope it gets up to bat and at least pushes a ground rule double.

Same for CPUs, a 3rd major CPU maker would be good to have as well.

[kl0012]

there is post there saying this 'Anyways I've read that SGEMM on AMD/ATI 5870 has reached 67 % of theoretical performance.

This would mean that the 5870 is doing 0.67*2.72 TFLOPs ~= 1.8 TFLOPs on SGEMM. Almost twice the peformance of LRB.'

don't know if its true or note though

I wonder why there are no official numbers on RV770/870 GPGPU perf in various realworld test (such as SGEMM, DGEMM). After all, AMD positioning GPU as application accelerator. As I understand AMD ACML supports matrix multiplication on GPU but the perf is pretty low (one says 300 GFLOPS on RV770). There are some highly specialized algorithms to extract more flops, but it is not clear if those algorithms are universal enough to use it in the real world.

[ajaidev]

I wonder why there are no official numbers on RV770/870 GPGPU perf in various realworld test (such as SGEMM, DGEMM). After all, AMD positioning GPU as application accelerator. As I understand AMD ACML supports matrix multiplication on GPU but the perf is pretty low (one says 300 GFLOPS on RV770). There are some highly specialized algorithms to extract more flops, but it is not clear if those algorithms are universal enough to use it in the real world.

If we take the 300 GFLOPS on RV770 on face value the RV870 maybe able to perform at-least 600 GFLOPS and thus the 2*RV870 "5970" can do almost 1200 GFLOPS.

When LRB comes out later, a RV970 maybe waiting and if AMD's current trend continues rv970 may as well have equal amounts of shaders as the 2*RV870 "5970" and thus may produce 1200 GFLOPS well at-least more than 1000 Gflops

[kl0012]

If we take the 300 GFLOPS on RV770 on face value the RV870 maybe able to perform at-least 600 GFLOPS and thus the 2*RV870 "5970" can do almost 1200 GFLOPS.

When LRB comes out later, a RV970 maybe waiting and if AMD's current trend continues rv970 may as well have equal amounts of shaders as the 2*RV870 "5970" and thus may produce 1200 GFLOPS well at-least more than 1000 Gflops

Isn't evergreen based firestream going to be released somewhere in Q2 2010? And assuming how Intel is determined to push LRB we can relay on 45nm. LRB with >32 cores very soon. But this is about GPGPU. Gaming perf is another story of cause.

[R101]

I want Windows that can use LRB cores as CPU ones!

The gains that would bring to the table would be

[eric66]

well .. on the economic side ... midrange gpu is a lucrative bite
by just reducing both ATi and Nvidia sales by 20% each ... it can cause quite a serious damage to them. most users purchase midrange gpus u know

well .. imho it could cause both ATi and Nvidia to end up selling highend gpus ...
but since we all know that intel has the tendency to "encourage" mobo manufacturers to build what they want (or else)

most probrably we can see a great shift in this?
the flaming gpu wars ... lol

well ati likes mid range products also so intel must first beat atis mid range products in order to get that share its not that easy. it took 2 succesfull generations for ati to steal only 4 percent or something

[Coldon]

two pages of arguing and speculation over a single result from a companies own presentation on a single "optimised" benchmark. I wouldn't put too much stock in the number just yet, <sarcasm> cause we all know intel would never optimize a score or misrepresent performance numbers</sarcasm>

yes, it supposedly outperforms current GPUs on a single synthetic test (wow it can multiply matrices together, OMG), once you show me some real world results doing something useful then I'll be impressed. If ATI/Nvidia had a presentation where they gave you a 3dmark score 50% higher than a competitor would you be impressed? I doubt it since it doesn't really represent real world performance and the score can easily be manipulated, so why are people so impressed now?

also the fact is that SGEMM is single precision, I'm curious to see double precision figures as well as the implementation of the SGEMM on the various GPUS. Also remember that the benchmark is being run on the various cards with different implementation and through different interfaces, so its not exactly a good test, if all three implementations ran through openCL on all the cards then it would be a better test. Again speculation and headlines based on a marketing presentation. *sigh...

I'm not biased to either camp, and i think larrabee would be a welcome addition to the market, but do i think its gonna be this super card that revolutionizes the industry, not a chance in hell...

[zalbard]

150w?

not bad considering that's less than a 5850 and it will be in a league of it's own in GPGPU.

Yeah, but it's 32nm, isn't it?

[Nedjo]

does anyone knows when does LBR is planned to launch? I mean some kind of solid info?

[Calmatory]

It is X86 so softwares should work pretty soon.

Let me guess, you aren't a programmer. Right?

[Coldon]

Let me guess, you aren't a programmer. Right?

+10000

[WaterFlex]

wow, thats totally insane chip! What is the future? more cores or less nm ? or both?

[Helloworld_98]

Yeah, but it's 32nm, isn't it?

this one's 45nm

[Olivon]

this one's 45nm

Right

A 32 nm final product would be nice for Intel

But no leaks about DX performance and drivers stability ... Recent newz about a possible launch on mid-2010 looks a little bit suspicious IMHO

[marten_larsson]

this one's 45nm

But isn't final product supposed to be 32nm and it is at that node that LRB is a 150W TDP card?

[kl0012]

does anyone knows when does LBR is planned to launch? I mean some kind of solid info?

Not a solid info but at least something:
http://www.brightsideofnews.com/news...n-1h-2010.aspx
late first half of 2010

Yeah, but it's 32nm, isn't it?

As far as i know the first LRB version was made with 65 nm. tech. Yet i don't know what version was shown by intel.