Planet Riva heeft een interview gepost met Michael Hara van nVidia. Er zit best wel interessante info tussen, bijv. dit stukje over Transformation & Lighting, een veelbesproken feature van next-generation videokaarten zoals de nVidia NV10 en S3 GX4:
Michael Hara : Just to be clear, we have not yet announced specifically which future products will incorporate Transformation and Lighting. We haven’t announced any new products since the RIVA TNT2, so I can’t make any comments to the product you refer to as “NV10.”
Transformation and Lighting are the first two of the four major steps (triangle setup and rendering being the final two) in the 3D graphics pipeline, or series of steps required to produce 3D graphics. Both steps are mathematically intensive and very specific in the mathematics they require.
The role of the transformation engine is to convert data that is described in World Space coordinates to Eye Space and Screen Space. World Space describes the 3D world, Eye Space describes the lighting and culling, and Screen Space describes the scene in the graphics frame buffer. Since each of these spaces use different coordinate systems, the 3D data must be converted or “transformed” from one space to another as it moves through the 3D pipeline.
The role of the lighting engine is to calculate the direction and distance from lights to objects in the 3D scene as well as the direction and distance from the objects to the viewer’s eye. The lighting engine also separates the distance information from the direction information to simplify future steps in the 3D pipeline.
Dedicated transformation and lighting will have the same impact as filtering did for pixels, but much bigger. Software transformation and lighting done by the host CPU today is the biggest bottleneck that has forced software developers to limit the geometric complexity and lighting sophistication of their 3D characters and environments.
Verder nog wat spul over texture compressing, bump mapping etc. Check 't hier.