Ahh the lost child. Wild to finally see it getting out into the public. It's crazy to see how much effort was put into it and then just went to /dev/null. For reasons, of course, but still insane.

tehsiggi wrote on Today, 11:20:

Ahh the lost child. Wild to finally see it getting out into the public. It's crazy to see how much effort was put into it and then just went to /dev/null. For reasons, of course, but still insane.

I mean a lot of the research they did for this went into other things so not all of it was a loss.

Trashbytes wrote on Today, 13:20:

tehsiggi wrote on Today, 11:20:

Ahh the lost child. Wild to finally see it getting out into the public. It's crazy to see how much effort was put into it and then just went to /dev/null. For reasons, of course, but still insane.

I mean a lot of the research they did for this went into other things so not all of it was a loss.

Well, some of it lived on in Xeon Phi, but to be honest, the whole graphics part of it was basically dumped. Which was a shame from the "novel concept" side of things.

Larrabee was cool. The idea was that you can optimize the GPGPU components, drop all fixed function components, and then your graphics chip was as versatile as it could be. I thought it still had some fixed pipelines, wikipedia says they kept the TMU.

I don't know if these would have been competitive in the games space. Maybe today, definitely not in 2010.

I remember going to an HPC conference in 2011 where they showed off Xeon Phi. I think Phi was more or less Larrabee without any of the graphics components. Phi had decent uptake, they had a ton of throughput, but they never displaced bigger and bigger graphics cards doing similar work. They likely had their niche but Intel could not scale it like graphics chip makers could.