Monday, May 18th 2009
Larrabee Only by 2010
Last week, Intel announced its Visual Computing Research Center at Saarland University in Saarbrücken, Germany. During its opening ceremony, details emerged about when Intel plans to commercially introduce Larrabee, the company's take on graphics processing using x86-based parallelism. The company categorically stated that one could expect Larrabee to be out only by early 2010.
"I would expect volume introduction of this product to be early next year," said Intel chief executive Paul Otellini. Until now, Larrabee was known to be introduced coarsely around the 2009-2010 time-frame. "We always said it would launch in the 2009/2010 timeframe," said Intel spokesperson Nick Knupffer in an email to PC Magazine. "We are narrowing that timeframe. Larrabee is healthy and in our labs right now. There will be multiple versions of Larrabee over time. We are not releasing additional details at this time," he added. In the same event, Intel displayed a company slide with a die-shot of Larrabee, revealing what looked like the x86 processing elements. Sections of the media were abuzz with inferences drawn on the die-shot, some saying that it featured as many as 32 processing elements.
Source:
PC Magazine
"I would expect volume introduction of this product to be early next year," said Intel chief executive Paul Otellini. Until now, Larrabee was known to be introduced coarsely around the 2009-2010 time-frame. "We always said it would launch in the 2009/2010 timeframe," said Intel spokesperson Nick Knupffer in an email to PC Magazine. "We are narrowing that timeframe. Larrabee is healthy and in our labs right now. There will be multiple versions of Larrabee over time. We are not releasing additional details at this time," he added. In the same event, Intel displayed a company slide with a die-shot of Larrabee, revealing what looked like the x86 processing elements. Sections of the media were abuzz with inferences drawn on the die-shot, some saying that it featured as many as 32 processing elements.
27 Comments on Larrabee Only by 2010
That means they're competing with next-gen Nvidia and ATI products... it'll be a fun period :)
Although, you don't need to be the performance leader to have profit (AMD i.e.).
I hope Larrabee doesnt turn out to be a flop, its such a unique architecture, Intels taking quite a gamble and has to fight the stigma their intergrated poopy graphics offering has left them with.
Just reading about Larrabee makes it sound ever more so awesome, built of the Pentium architecture but modified like crazy, the way it splits threads into strands and all that is cool it, doesnt use stream processors afaik.
I think PC Perspective have some detailed articles about it.I believe that to begin with, we'll be lucky if we see close to the same peformance of current cards, the benefit is though that unlike the card manufacturers which will have to go back to the drawing boards when slapping more SP's on doesnt make their cards faster, Intel (provided the product doesnt get canned) could in theory, just cruse along with their scalable peformance, not needing to worry about hitting some peformance wall.
Anyway, Intel to face the same conundrum they had with Itanium which I hope doesnt stop larrabee from being a big change to computing/gaming/etc.
With 32x 512-bit vector MIMD it will whoopass CUDA.
512-bit can handle 8 double precision (32 bit) or long integer variables. And 32x. That is equivalent to 256x the speed of a regular math unit in a single core processor.
That means for math work, Larrabee has the potential to be (256/4)=64x as fast as a quad core CPU clock-for-clock. It will probably be running at half the frequency of a regular CPU, maybe slower. But that will equate to at least 16-24x the speed of a quad core.
And math libraries can be very rapidly modified for Larrabee x86.
R.I.P. CUDA/Tesla.
nV know this. GT300 has to compete. Look at the panic in their eyes.
Moore's Law will be bust open for the 2008-2010 season. And the Soda-Darwin law applies: survival of the fastest.
well, maybe not lol:D
but its really funny how nobody is talking about how Nvidia stole Intels plan to use MIMD, they obviously did, but everyone still busses on Intel for doing things different
what the hell does the "only" in the title actually mean?
I think he meant"what's their next move in the graphics sector?"
I do believe Intel can make a better one eventually, but they need to make the decision to put the effort and time in for the long term.
Again, Intel's design is superscaler. It isn't very hard for them to release a processor with far more cores but it comes at a price. Intel shouldn't have much problem staying way ahead of NVIDIA and AMD.
Remember, the only thing that's really special about GPUs is there extremely high FlOp performance. It doesn't matter how a GPU gets (only reflected in price/profit margins) it so long as it gets it.