Intel drops plans for Larrabee GPU in 2010
Intel has scrapped plans to release its Larrabee many-core graphics processor next year. The chipmaker has unveiled plans to use the technology for a software development platform instead.
The prospect of a standalone Larrabee GPU being released at some point is not out of the question, but such a chip will not be the first Larrabee product, due to delays in silicon and software development, Intel said in a statement sent to ZDNet UK on Monday.
"Larrabee silicon and software development are behind where we had hoped to be at this point in the project," the statement read. "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."
This follows Intel Labs' release of another many-core chip called SCC on 2 December. Like Larrabee, SCC is based on the x86 architecture but with different interchip communications. It is also aimed at internal and external software development.
The chipmaker said it intended to create a high-performance computing (HPC) platform based on Larrabee, following an "extremely promising" demonstration of such applications at the SC09 conference in November.
"While we are disappointed that the product is not yet where we expected, we remain committed to delivering world-class many-core graphics products to our customers," Intel said. "Additional plans for discrete graphics products will be discussed some time in 2010, and our plans to deliver the world's first CPUs with integrated graphics this month are unchanged."
Intel first announced Larrabee in 2007, touting the many-core architecture as being aimed at HPC. The company fleshed out the details in 2008, when it said the first Larrabee product would be a graphics processor using x86 cores, achieving competitive performance through ray tracing rather than the polygon manipulation used by existing high-performance graphics devices.
Gartner analyst Steve Kleynhans told ZDNet UK on Monday that Larrabee was, in his view, always going to be "a relatively slow ramp".
"The product that Intel wanted to bring out initially wasn't going to have the kind of performance needed to compete against [currently available] high-end discrete graphics products," Kleynhans said, adding that "the worst thing Intel could do is bring out a product that would end up disappointing to a level that it tainted the brand".
According to Kleynhans, the first one or two Larrabee products were unlikely to be big moneymakers for Intel anyway.
"There have always been two sides to Larrabee," Kleynhans said. "One was as a new discrete graphics card, but probably the more important side was as the many-core compute engine."
"It's still available for anyone who wants to develop many-core applications, and we're still at a very early stage with that whole wave of software development, so having a development tool rather than a mass-market graphics product is probably quite appropriate [for] the real long-term goal of the whole Larrabee project."