Is Larrabee Dead?: The Tangled Future of CPUs and GPUs

December 11, 2009 -By Michael J. Miller
Is Larrabee Dead?: The Tangled Future of CPUs and GPUs

http://tinyurl.com/ydfglks

We've seen a lot of stories this week about Intel's announcement that it
would not be shipping a version of its "Larrabee" processor as a
graphics chip in the next few months, as the company had previously
announced. Instead, Intel says it will ship this processor only as a
development platform for now, while continuing to work on it. While some
have taken this as the "death" of Larrabee, I see it mostly as a
continuing indication of just how hard it is to bring the theory of
parallel processing into the practice of the real world.

I'm certain Intel will continue to work on many of the concepts within
Larrabee, because the overall directions of both the CPU and the GPU
markets require it. As I see it, the two big trends in general processor
designs are a push for much greater parallelism, particularly for
high-end computing; and much more integration, particularly for
mainstream and low-end computing. The Larrabee project had implications
for both of those; and for how Intel, AMD, and Nvidia compete going
forward.

Intel Guy wrote:
December 11, 2009 -By Michael J. Miller
Is Larrabee Dead?: The Tangled Future of CPUs and GPUs


I think the only way Intel is going to finally be successful at GPUs is
when they decide to follow the lead of the market, rather than trying to
dictate to it. This is relevant to most of their other ventures too, not
just GPUs.

Yousuf Khan