.
/v3-uk/news/1968429/intel-sheds-light-larrabee
05 Aug 2008, Shaun Nichols , V3
Intel is revealing new details on its forthcoming Larrabee chip.
The new graphics processor is Intel's foray into the high-end graphics market currently dominated by nVidia and ATI.
Intel revealed that the first of the Larrabee offerings is scheduled to hit the streets in 2009 or 2010. The company initially plans to use the chips in discrete graphics cards and support both OpenGL and DirectX.
In the long term, however, Intel also sees Larrabee influencing the CPU market, which is increasingly moving towards the multi-core processing model used in GPU chips. ATI and nVidia have similar efforts in place.
"Larrabee is expected to kick start an industry-wide effort to create and optimise software for the dozens, hundreds and thousands of cores expected to power future computers," the company said.
"Over time, the consistency of Intel architecture and thus developer freedom afforded by the Larrabee architecture will bring about massive innovation in many areas and market segments."
Intel plans to differentiate Larrabee through the use of its x86 architecture. Larrabee will be based on the Pentium processor.
The company did not say how many cores would be offered in the chip, but did say that each of Larrabee's cores would be able to handle four separate execution threads and will borrow some of its pipeline components from the dual-issue Pentium chip.
Intel plans to release more details on the Larrabee next Tuesday in a paper to be released at the Siggraph conference in Los Angeles.
Do you agree?
ray tracing doesn't need more compute
Adding more multi-core compute is not what we need.
Current trends in microprocessor evolution indicate that more and more applications will become memory-bound and be unable to benefit from the ample compute resources offered by the next generation of computers.
This increasing disparity between a system?s compute capabilities and the available memory bandwidth needed to fully utilize that compute is the real problem.
Until the memory bandwidth issues are solved, adding more compute won't make a dent at speeding up performance for applications like ray tracing - that's the real bottleneck that needs to be addressed.
Posted by AndyZee, 06 Aug 2008