Fill Rate on GeForce 256

How many megapixels was that?

Yes, you heard correct the first time. The GeForce 256 can push out a sustained fill rate of 480 million pixels per second. That's over double the fill rate of the original TNT2. NVIDIA accomplishes this through GeForce's four-pixel pipeline rendering engine. We can derive the core speed of the GeForce 256 by dividing the pixel fill rate by the 4 pixel pipeline (480/4 = 120). Apparently, the GeForce 256 runs at a graphics core speed of 120MHz. By implementing massively parallel pipelines, NVIDIA is able to greatly increase the pixel fill rate without increasing clock speed. In fact, this speed is 5MHz lower than the TNT2 vanilla's core speed of 125MHz.

So….what about…megatexels?

What's really interesting here is NVIDIA's lack of confirmation on texel fillrate. From this, I think it's safe to assume that we're not going to see 2 textures per cycle on each of the pixel pipelines (which would have added up to a consummately reportable 960 million texels, enough to take on the spec-sheet wonders of BitBoys' Glaze3D).

Instead, what should we expect? Most likely, 1 texture per cycle per pipeline, or an equivalent of such, giving us 480 million texels per second, only modestly higher than the nearest competition, especially in light of their "order of magnitude" improvements brought forth by onboard T&L. Without stretching too far, if the GeForce is limited to 480 Megatexels/sec, we have no doubt that 3dfx's upcoming fall 99 product will be able to claim first place in that category.

Where's the beef?

If we have a beef with the GeForce 256, it's certainly the clock speed. With over 23 million transistors, the GeForce exceeds Pentium III or Athlon in complexity, and the limitations of the AGP slot and power supply are already being felt. Being able to scale in parallel is a great alternative, but the fact remains that every title benefits from faster clock rate, and in many situations, there is no substitute.