Larrabee Failed

[Hitman.1stGame]

In our extensive coverage of Intel's project Larrabee, one question persisted: is it worth for Intel to invest billions of dollars on a market that cannot be dominated? Intel Larrabee as a desktop card for consumers - is dead. As a consequence of a chain of events, Intel's executive management decided to stop pouring millions of dollars in a bird that failed to fly.

As of Friday, December 4, 2009 Intel decided to stop investing into Larrabee as a consumer project. In a statement given to us by Nick Knupffer, who worked as Intel's spokesman for Larrabee, it was stated that "Larrabee silicon and software development are behind where we had hoped to
be at this point in the project. As a result, our first Larrabee product will not be launched as a standalone discrete graphics product, but rather be used as a software development platform for internal and external use."

Source

 

[Wile E]

This is actually a bummer. Although I doubt it would've had 3d performance approaching the ATI and nV camps, it had a lot of potential as a co-processor.

 

[HossHuge]

I wonder if this has anythng to do with the huge fines they have had to pay recently?

 

[FordGT90Concept]

I doubt graphics on Larrabee is completely dead, it is just delayed. Coding all of the DirectX and OpenGL platforms in to a driver will take a lot of work for them--more than making the hardware capable of executing x86 instructions. The question is, will the Larrabee cards that do come out have video out ports? If yes, a driver update in a year or so will turn those co-processor cards into GPUs.


It also sounds like the first round is intended for driver development which is no surprise.

 

[Fourstaff]

I was hoping that you can play games with Intel GMA 1000000 :shadedshu

 

[lemonadesoda]

I think Intel will shift focus for larrabee from graphics to HPC and CUDA/OpenCL stuff.

The larrabee cores are so much more powerful and adaptable than just running graphics draw and shading. In a sense, they are overkill. Underperforming in graphics, and using only "10%" of their potential. Wrong tool, wrong application.

Stick a couple of larrabee 64-core cards into a workstation and you will have a miniframe when it comes to crunch time. IMO, that is the way to go. Rather than small incremental CPU improvements, just add whallop to a workstation with a drop-in card. A bit like clearspeed, only bigger, faster, and with more marketing and development $.

 

[KainXS]

they said this before already, its not gonna be a consumer card, its gonna be for HPC market, im still dissapointed about that, . . . . . .

 

[Steevo]

This on the tail of their payout to ATI....mebey some undisclosed agreements.

Use larrabee tech on stream processors to accelerate programs, and I still have the idea that ATI gave up some GPU tech to Intel to help create a push against Nvidia.

 

[Nailezs]

what a bummer, if nothing else i was looking forward to the competition

 

[cadaveca]

Wow. I was not expecting this, really.

our first Larrabee product will not be launched as a standalone discrete graphics product, but rather be used as a software development platform

But this tells me Larrabee is not dead. Just in it's current form. they will take what they have learned, and make a new product.

 

[Hitman.1stGame]

CPU vs GPU

I think Intel will Never have powerful GPU if they still depending on Larrabee architecture .
simply because Larrabee are just toomany small cores of weak X86 processors OK
and if we compare CPU run at 3GHz Like E8400 against any modern GPU what score will have on chart
1:100 or something ?
i will make it simply see ur CPU performance On 3dmark CPU test
u will see ur CPU suffering against poor graphic and low resolution scenes .

 

[Yukikaze]

I think Intel will Never have powerful GPU if they still depending on Larrabee architecture .
simply because Larrabee are just toomany small cores of weak X86 processors OK
and if we compare CPU run at 3GHz Like E8400 against any modern GPU what score will have on chart
1:100 or something ?
i will make it simply see ur CPU performance On 3dmark CPU test
u will see ur CPU suffering against poor graphic and low resolution scenes .

Using x86 cores is not a bad idea for a GPGPU clone/HPC co-proc, where the concept will shine. The question is whether it will ever make it to a consumer card in the end. I suspect it will be limited to the HPC market, but we might yet be surprised.

BTW, that rush of wind you just felt was the collective sigh of relief from every single nVidia employee

 

[wahdangun]

yeah, intel always have crap driver suport, remember they can't even use hardware TnL, what a joke

 

[i789]

Larrabee is just delayed for the consumer market, it is still going get released next year for the developers.

 

[Binge]

Larrabee is just delayed for the consumer market, it is still going get released next year for the developers.

I don't think so Tim.

 

[Lionheart]

Wow, this is a bummer and good news at the same time

 

[DrPepper]

I think Intel will shift focus for larrabee from graphics to HPC and CUDA/OpenCL stuff.

The larrabee cores are so much more powerful and adaptable than just running graphics draw and shading. In a sense, they are overkill. Underperforming in graphics, and using only "10%" of their potential. Wrong tool, wrong application.

Stick a couple of larrabee 64-core cards into a workstation and you will have a miniframe when it comes to crunch time. IMO, that is the way to go. Rather than small incremental CPU improvements, just add whallop to a workstation with a drop-in card. A bit like clearspeed, only bigger, faster, and with more marketing and development $.

That sounds like the most logical idea I've seen in this thread.

 

[Apocolypse007]

i don't see as how this is good news. I always welcome competition and new approaches to hardware development.

I kind of saw this coming however. Larabee only had a shot at mainstream/mid performance markets if it were on schedule. The delays only made it uncompetitive at the point of release. As such, it is more useful now as a co-processor than a GPU. If I were an Intel executive I would have done the same thing. Why release a product that took millions to develop into a market where it cannot compete?

 

[jagd]

Im surprised noone mentioned legal wars
"Intel filed a suit against nVidia in Delaware court [naturally, since both companies are incorporated in the "Venture Capital of the World" state], claiming that nVidia doesn't hold the license for CPUs that have integrated memory controller. nVidia didn't stand back, but pulled a counter-suit, but this time around, nVidia wanted the cross-license deal annulled and to stop Intel from shipping products that use nVidia patents.

If you wonder why this cross-license agreement is of key importance for Larrabee, the reason is simple: without nVidia patents, there is no Larrabee. There are no integrated chipsets either, since they would infringe nVidia's patents as well. "

http://www.brightsideofnews.com/new...h-intel-larrabee-story-revealed.aspx?pageid=2

Performance or legal (patents ) problems would be nice to know truth ,btw this is unbeliveable intel and nvidia trying to destroy each other at court but everyone closing their eyes.

 

[FordGT90Concept]

What part of Larrabee would require an NVIDIA patent? If Intel wanted an SLI-like technology, they'd buyout Lucid.

 

[Zubasa]

What part of Larrabee would require an NVIDIA patent? If Intel wanted an SLI-like technology, they'd buyout Lucid.

Or even easier, just go bargin with AMD.
They can even develop their own, since they basically have all the money in the world.

 

[human_error]

The reason it was cancelled (note it is only the current consumer chip design binned, not the R&D into future chips) was because it was weak. It had around 1 TFLOP of power - a 5870 has 2.7 TFLOPS - and larrabee was 6 months away from launching as external companies hadn't even started sampling.

This was a good decision for Intel - if sold the current larrabee would have to have been in the low-mid range segment (against fermi and tweaked HD5k series chips it would have to be sold in that segment). It would also have had incompatability issues and some driver bugs (as new platforms always do) so most people who would buy it would get burned by these issues and intel's discrete card reputation would be awful unless intel had to invest a ton more money into fixing things quickly, but that would just be a waste. A bad reputation or unnecissary costs is not what intel want, especially since they'd make very little money from it and have plenty of spare cash anyway.

Intel working on the software side primarily seems to indicate they will be making scalable drivers which will be able to work on future dedicated cards with tons or cores (which they'll still be working on making competative) and also on intel cpu's (maybe with a little dedicated silicon for the drivers to run on too in future designs) - this is the best solution for intel to work on at the moment as suddenly all intel cpu machines could have dx11/dx12 capability - yes it would be rubbish performance on current cpu's but it would be good for marketing to have, and if intel could get something out which competes with ati/nvidia offerings then they could sell the hardware as well.

 

[jagd]

I dont know what part need nvidia patent but while nvidia has 1000 patents it is hard to not violate one (you can ask to writer ,i just queted) http://blogs.nvidia.com/ntersect/2009/11/nvidia-celebrates-1000th-patent-1.html . intel already bought lucid shares http://www.intel.com/capital/news/releases/070110.htm

What part of Larrabee would require an NVIDIA patent? If Intel wanted an SLI-like technology, they'd buyout Lucid.