Pop quiz hotshot. You are already manufacturing the fastest 3D accelerator on the planet. You are breaking into new markets previously dominated by your biggest competitor, and you are looking to once again redefine the way tomorrow's games are developed and played. What do you do? What do you do?
Here's a hint: take 57 million transistors, 4 pixel pipelines with two texture units per pipeline, and stir in the DirectX 8 specification. As an added measure, make that pixel pipeline fully programmable and include a hearty library of ready-bake vertex processor effects to make things easy for your neighborhood game developer. Polish up that FSAA and make sure you do not forget legacy support for those DirectX 7 games that have yet to emerge!
If this sounds like a walk in the park, then NVIDIA may very well have a job for you. However, if you are like the rest of us, interpreting the GeForce3's list of brand new features is about as easy as whipping up a batch of Blintzes from a Yiddish cookbook. Not to worry though - we are here to explain the technology behind the GeForce3 and how it will affect the DirectX 8 applications you buy later this year.
According to NVIDIA, processing graphics takes raw transistors - there is just no way to skirt the issue. NVIDIA's RIVA 128 processor was comprised of 7 million transistors. The TNT2 crammed 15 million transistors onto a single die, while the GeForce2 upped that number to 25 million. For the sake of comparison, ATI's RADEON is made of 30 million, and Intel's new Pentium 4 processor houses 42 million transistors on a .18-micron manufacturing process!
For the GeForce3, NVIDIA was able to integrate no less than 57 million transistors onto a single die manufactured on TMSC's new .15-micron process. This new logic includes legacy support for static T&L, which will be important for compatibility with DirectX 7 applications.