1. Take a GeForce2 GTS GPU, cook with TSMC's .18 micron GX process.
2. Set the GPU clock to 250MHz.
3. Top the GPU with a heat sink and fan combo.
3. Pour in 64MB of DDR SDRAM.
4. Stir the DDR SDRAM to 230MHz DDR, or the equivalent of 460MHz.
5. Garnish the memory with heat sinks.
6. Let it sit overnight.
7. Enjoy.

Yes, that's right, a 250MHz core clock matched to 230MHz DDR SDRAM. The GeForce2 Ultra's core is the same as the GeForce2 GTS', so it contains four pixel pipelines each capable of rendering one pixel per clock with two textures per pixel. Four pixels per clock at 250MHz gives you a one gigapixel per second fill rate. And at two textures per pixel, we're talking a two gigagtexel per second texel rate. Ladies and gentlemen, that's fast! TSMC's .18 micron GX process is a newer process technology, which apparently enables NVIDIA to push their GeForce2 architecture a bit further.

And backing up the blazing fast GPU core is 64MB of 200MHz DDR SDRAM. According to NVIDIA, that's the fastest DDR SDRAM memory available on a consumer product today. With the equivalent of 460MHz memory running on a 128-bit memory bus, the GeForce2 Ultra carries an astounding 7.4 gigabytes per second of memory bandwidth. That's faster than a four PC800 RDRAM channel system! With such a stunning amount of memory bandwidth, the GeForce2 Ultra is sure to run like no other card before it at 1600x1200 and beyond.

And we cannot forget the hardware transformation and lighting engine. At 250MHz, the GeForce2 Ultra GPU can push 31 million polygons per second, compared to the GeForce2 GTS' 25 million polygon per second rate. Considering that so few games today make good use, let alone any use, of the polygon pushing power of the GeForce 256 and it's 15 million polygons per second rate, 31 million polygons per second is just overkill.

Otherwise, beyond the major speed improvements, the GeForce2 Ultra architecture is essentially the same as a GeForce2 GTS.