bit-tech.net

NVIDIA talks G80

NVIDIA talks G80

Based on what NVIDIA has been saying, it looks like G80 is not going to be a unified shader architecture.

An NVIDIA representative has given some interesting tidbits on the company's next-generation G80 graphics processing unit at the Morgan Stanley Semiconductor & Systems Conference.

The spokesperson stated that they're "increasing the flexibility of the programmability, enabling the artists to express themselves in a free way." That doesn't tell us a great deal about their hardware configuration, but they have stated that programmers will be able to express themselves by programming in a unified shader pipeline.

Obviously, DirectX 10 is going to change things somewhat, as it completely unifies the graphics pipeline. However, we get the impression that NVIDIA's G80 architecture will be a "hybrid" design, based on the discussions we had with NVIDIA's Chief Scientist, David Kirk.

Unifying the pipeline has some distinct benefits, but it can also have some drawbacks as Kirk mentioned in our interview last year. "Another word for 'unified' is 'shared', and another word for 'shared' is 'competing'." Based on this, we feel that NVIDIA will handle the unified API at driver level, rather than in hardware. However, we understand that ATI's architecture will be completely unified at the hardware level.

The stated that NVIDIA has been working on their next generation architecture since 2002, and they've invested around $250M into it already. By the time it launches later this year, they will have invested close to $500M into research and development for their upcoming architecture. You can read more here.

Who do you think will be the first to release a DirectX 10 GPU? Discuss in the forums...