3dfx and the future
If Savage2000 does 700Mtexels per second (2 textures on 2 pipelines at 175MHz), and GeForce 256 does 940, what do you think we can see from their competition? What would you say would be a good fill rate number for a next-generation card?
I am not sure where you get that the GeForce can process 940 Mtexels/sec. Well, I guess if you count trilinear filtering as 2 "texels" then the 940 number is accurate (since GeForce performs trilinear in a single clock in every raster pipeline). But this number is very misleading because it implies that these are separate texture maps, whereas they are simply trilinear filtered. I wouldn't get too worked up about Marketing performance numbers - real world performance is the only thing that matters at the end of the day…
Is 3dfx exploring texture compression for its next-generation products?
Your timing is impeccable, Kenn. Very perceptive. There is a misconception that we do not support texture compression now. We have, in fact, had texture compression in all 3dfx graphics processors since the original Voodoo Graphics. We have some very exciting developments happening in this area and plan to talk to you more about that very subject next week.
What size manufacturing process is 3dfx looking at for future products?
I think everyone knows that 3D graphics accelerators are now right up there with CPUs in terms of pushing the silicon lithography limits. We'll have both .25/.22 and .18 micron products in the very near future.
Is 3dfx considering embedded DRAM?
We constantly evaluate all memory options, both external and embedded. Nothing more to say about it right now, though.
What is the next 3dfx product going to be called?
How about "3dfx Next Generation"? NOT!
How about just a chipset name?
…see answer to question #13…
Is this product going to be faster than GeForce 256?
As I mentioned previously, there are numerous performance characteristics which interact to form the sustainable performance of a 3D accelerator. Nvidia has focused heavily on the geometry bottleneck with the GeForce 256, almost at the exclusion of substantially increasing fill-rate performance. In the very near term, we have gone down almost exactly the opposite route, focusing much more heavily on fill-rate performance. As a result, as gamers want to run their games at high resolutions and 32bpp color depth, we believe hands down we will have a faster product than GeForce 256 across the majority of games.
There were some misconceptions that the T-Buffer was the only trick 3dfx had up its sleeves. Anything you'd like to say about that?
Give us a little credit please. 3dfx has been at the forefront of 3D technology ever since the original Voodoo Graphics. T-Buffer is a very exciting technological breakthrough, but we've just gotten started with a host of extremely compelling new features in the future. We're going to leave people's heads spinning with incredible special effects engines that will be productized in the near future.
What else is 3dfx looking towards to implement "Hollywood on the desktop?"
We can't really comment at this time about what's next after the T-Buffer. The T-buffer for the first time allows game producers to have many of the same tools previously only available to motion picture directors and cinematographers. But there are many, many new effects and capabilities that we'll be announcing in the future.