Exclusive: Intel Larrabee 5 times faster than Nvidia, AMD

[NeedMoMegaHurtZ]

I hope it's even faster.

That way it will put some pressure on the Red and Green Teams to develop some nice GPUs (instead of just glueing some together or renaming some )

[JumpingJack]

any bases for this? couse i'm a bit tired of reading this articles with just speculation.

Also remember Larrabee is way off and will compete with something completely diferent (much better) that the current line up from NVIDIA or ATI

They base it off of theoretical calculations achievable with the width of the operands, the number of executions per cycle, and the number of course as a particular clock speed.... until there is real silicon, it is purely 'based on simulation' like information.

[clayton]

I thought they meant 5x faster than AMD/Nvidia's best until I read IGP. Maybe by 2016 it'll catch up to the 8000 and 3000 series.

[JumpingJack]

I thought they meant 5x faster than AMD/Nvidia's best until I read IGP. Maybe by 2016 it'll catch up to the 8000 and 3000 series.

Larrabee is not an IGP related product, it is going into discrete graphics cards as well as HPC accelerated sockets.

[Lestat]

so how would something based off a VERY old cpu that was not that good be better than today's cpu's. based on an MMX cpu.. pfpfpf

listen boys and girls Intel is a CPU maker, GPU making is VERY VERY different and if you have ever tried intels GPU's, you know, for a fACT, for a 100% freaking holy jesus FACT that intel is NOT, never has been and never will be a GPU maker.
you cant just start making GPU's overnight.

i say this is horse crap, and yes that is my opinion, but if you read this forum and many MANY others you find that this is the same opion of so many others.

I dont think we doubted the Core series cpu's but this isnt a cpu, its completely different. and for Intel to say "we're basing this GPU, which isnt even out market, off an old cpu that was not very good, then i say they are just higher than a kite to even insuate that the performance would be even to the level of today's gpu's.

[Macadamia]

<---------------------- I'm with Itanium

[WRC]

so they just sarted making SSD's overnight and look how good they perform. Unfortunately intel has the money and the engineers to build a gpu like this, however by the time the card is due for a release nvidia and amd will outperform the larabee

[Revv23]

Wow this will be interesting.

[hollo]

AMD:
3dmark score = 1.1 x nv
ingame fps = 0.9 x nv
Intel:
3dmark score = 2 x nv
ingame fps = nv/2

[dinos22]

Lestat
you are forgetting the sheer size of that company. They can just buy the experts and give them whatever they want to crush the competition. I wouldnt be surprised to see them going gungho and being very competitive by the time they hit 2/3rd gen of GPUs

[Zytek_Fan]

Lestat
you are forgetting the sheer size of that company. They can just buy the experts and give them whatever they want to crush the competition. I wouldnt be surprised to see them going gungho and being very competitive by the time they hit 2/3rd gen of GPUs

Which is why they pretty much hired anyone with experience in graphics before the Larrabee project was greenlighted

[Sparky]

And Larrabee will be alot more efficient in the resource usage than current GPUs.

Yeah but by 2010 that could easily change.

[[XC] gomeler]

so how would something based off a VERY old cpu that was not that good be better than today's cpu's. based on an MMX cpu.. pfpfpf

listen boys and girls Intel is a CPU maker, GPU making is VERY VERY different and if you have ever tried intels GPU's, you know, for a fACT, for a 100% freaking holy jesus FACT that intel is NOT, never has been and never will be a GPU maker.
you cant just start making GPU's overnight.

i say this is horse crap, and yes that is my opinion, but if you read this forum and many MANY others you find that this is the same opion of so many others.

I dont think we doubted the Core series cpu's but this isnt a cpu, its completely different. and for Intel to say "we're basing this GPU, which isnt even out market, off an old cpu that was not very good, then i say they are just higher than a kite to even insuate that the performance would be even to the level of today's gpu's.

You fail to understand the sheer size of Intel.. a few hundred people with a huge budget can create wonders. Don't underestimate what lots of people with lots of resources can do, a few years of research and modifications yielded Conroe

[ziddey]

how was the atom a huge success? man what a joke

[Drunner611]

And I'll be running the GTX 430 OMGWTFBBQFTWROFLLMAO Edition.

I doubt it. By that time the GTX 500 series will be out and your card
will become the GTX 430 OMGWTFBBQFTWROFLLMAO+ Edition

[shiznit93]

people that doubt intel forget that this is a completely vertically integrated company that basically turns sand into money as Nvidia well knows, with the best manufacturing tech in the industry and the abilty to hire/buy experts just as well as anyone else.

[G0ldBr1ck]

It seems the magic will be in the software, If they can pull that off then it will perform.

[003]

Larrabee hasn't shipped so you don't know what it is and I don't know what it is. That said, by the time it does ship, Nvidia's technology will be so far advanced it won't matter.

^^^

[Dami3n]

First they must learn how to build decent igp, because ALL of their igps are crap. When they get it, they can start to think in serious gpus.
Larrabee->hype->

[Blacky]

I would like to see this, but knowing how much intel fails on the graphic department well puts me on doubt if larrabe could be a success when it launches...and like some of here mentioned, by the time larrabe arrives nvidia & amd will have a more advanced, faster, more efficient gpu on the market. I could be wrong on larrabe tho, but time will tell...

[003]

Quad 4870s are capable of just under 5TFlops. 64 pentium MMX based cores at 2GHz, I doubt would even break 1TFlop (a heavily overclocked yorkfield doesn't even break 100Gflops). The only thing intel has going is the brand name.

[Etihtsarom]

Well, pointless to debate the merit of tech that's 2 yrs out. Logic dictates that if they do well, we win. If they don't do well, we have more competition on price and competition for better technology, we win again.

[JumpingJack]

Quad 4870s are capable of just under 5TFlops. 64 pentium MMX based cores at 2GHz, I doubt would even break 1TFlop (a heavily overclocked yorkfield doesn't even break 100Gflops). The only thing intel has going is the brand name.

http://softwarecommunity.intel.com/U...e_manycore.pdf

The theoretical throughput for a single larrabee core at 1 GHz is 32 GFLOPs, at 16 cores 1 GHz this comes in at 512 GFLOPs, running at 2 GHz would be 1 TFLOP, 64 running at 2 GHz would be 4 TFLOP, 64 core running at 2.5 GHz would be 5 TFLOP.

Your error is assuming Larrabee is simply a single Pentium MMX core ... this is too simplified of an assumption, Intel has constructed the core to process 4 threads with up to 16 SP vector operations per clock per core.

Jack

[metro.cl]

They base it off of theoretical calculations achievable with the width of the operands, the number of executions per cycle, and the number of course as a particular clock speed.... until there is real silicon, it is purely 'based on simulation' like information.

Cool but for me is still a rumor (this news) and way to far away (Larrabee), also there are so many variables when you want to enter a "new" market (in this case for Intel) that even the biggest guy in the world wouldnt claim to sweep the competition.

[JumpingJack]

Cool but for me is still a rumor (this news) and way to far away (Larrabee), also there are so many variables when you want to enter a "new" market (in this case for Intel) that even the biggest guy in the world wouldnt claim to sweep the competition.

I don't disagree ... on paper it sounds cool, conceptually interesting, potentially competitive ... but it is vaporware until actual product is shown and tested.