Larrabee cancelled

[Rollo]

http://www.semiaccurate.com/2009/12/...ture-variants/

It is SemiAccurate/Charlie, so definitely in rumor status for now.

[Rollo]

http://news.cnet.com/8301-13924_3-10...orsPicksArea.0

Confirmed on CNET. The L8trB crashes and burns, Intel's second foray into the discrete graphics is stillborn.

[BUFF]

according to anandtech it's only the initial generation planned for retail - Larrabee project is still ongoing.
http://anandtech.com/weblog/showpost.aspx?i=659

[Lee H]

From what I can guage from the information, the project lives on, just the current form of larabee is canned.

It's a shame really, it looked promising and would have forced other card manufacturers to bring out new GPU's instead of simply renaming their GPU cores every 6 months to another name

[circuitmonkey]

It makes me wonder why now that they've successfully shown that 16 x86 cores over a PCI-E bus can work as a GPU, and more recently 48 x86 cores can be put into a single CPU whey they'd proceed with slowing things down over the bus and not just cram as many cores on the CPU and use that for both CPU and GPU?

[aidanjt]

Because x86 is simply far too inefficient and expensive to cram a crapload of cores into a PCIe card (or even a CPU package) and call it a competitive GPU.

[Rollo]

I think Intel found out breaking into the high end GPU market is a bigger project than they thought. This was difficult for them for the same reason making a CPU would be difficult for ATi or NVIDIA- their core expertise and experience is elsewhere.

[aidanjt]

It's not that it's a project too big for Intel. It's that Intel engineers are being hamstrung by the idiotic x86-everywhere mentality of upper-management, or 'not invented here' syndrome.