NVIDIA's 60.72 drivers vindicated?
It looks like the most recent NVIDIA driver controversy may be much ado about nothing. Driverheaven discovered slight mip map level of detail differences between the GeForce 6800 Ultra with Forceware 60.72 drivers and Microsoft's DirectX 9 reference rasterizer in 3DMark03 and Max Payne 2. Since Futuremark had certified the 60.72 drivers as meeting its optimization guidelines there was some concern that NVIDIA was up to something. However, it appears that that the slight mip map LOD differences are due more to an inadequacy of the DX9 reference rasterizer than any questionable driver optimization on NVIDIA's part. Here's an excerpt from Futuremark's response to Driverheaven's findings:
The explanation for the rendering differences is GeForce 6800's new isotropic filtering algorithm: It differs from that of the reference rasterizer. There is an analysis of this available at Digit Life (http://www.digit-life.com/articles2...40-part1-d.html). Based on all information we have, it seems that this algorithm is applied consistently the same way regardless of the application, i.e. it is not application specific and it is not targeting 3DMark specifically.
We also contacted Microsoft about the difference between the rendering of the reference rasterizer and NV40: "The DX9 reference rasterizer does not produce an ideal result for level-of-detail computation for isotropic filtering, the algorithm used in NV40 produces a higher quality result. Remember, our API is constantly evolving as is graphics hardware, we will continue to improve all aspects of our API including the reference rasterizer."
Futuremark also goes on to say that NVIDIA's isotropic filtering algorithm has no impact on performance.