Intel at the opening of Saarland University's Visual Computing Institute on Wednesday gave clues as to its plans for its many-core Larrabee graphics processor. In a presentation to guests, the semiconductor firm showed that the design uses 32 processor cores and mates each of these with its own vector math unit, allowing each core to process very specialized tasks very efficiently. They also have a pool of cache memory and a memory interface on their edges.
Larrabee will combine a large array of Intel Architecture (i.e. x86) cores with a brand-new cache architecture, a new vector instruction set, and a new vector processing unit. Gelsinger specified that Larrabee's programmable architecture will allow it to accelerate anything from high-definition video and audio processing to physics, artificial intelligence, and global illumination. it's confirmed that Larrabee will be compatible with DirectX and OpenGL application programming interfaces, so Larrabee should be able to run existing games.
Additionally, Intel's Joseph Schultz mentioned at the event that it has moved back the launch of Larrabee from its original late-2009 target to the first half of 2010. The reasons for the delay weren't mentioned.
Larrabee is Intel's first relatively high-performance graphics design and is intended to overcome the flexibility problems of dedicated video chipsets. By using many x86-based cores, the company hopes to give developers a heavily programmable architecture that lets them handle custom visual effects as well as running many tasks at once.