The Future, T&L, VSA-100 Refresh

FiringSquad: How do you see 3D accelerators advancing in the future? Is clock rate going to continue to increase, or will manufacturers look elsewhere (perhaps to additional parallelism) for performance gains?

Yes, yes, and yes. Clock speeds will continue to increase, memory widths and speeds will get wider, and we'll continue to look elsewhere architecturally such as increasing parallelism and even more arcane things like chunking. The great thing about 3D on the PC and in consoles today is that we're nowhere close to enabling real-time cinematic realism on the desktop. We can easily see a need for five orders of magnitude higher fill rate than provided by Voodoo3 today. If Scott Sellers wants to, he can design better and better 3D chips for the next 30 years!

FiringSquad: Does 3dfx has a huge advertising campaign set for the launch of Voodoo4/5?

We will have a substantial and aggressive marketing campaign behind our new products. It will not have the same budget as we had when we launched the new company and the new products simultaneously with the release of Voodoo3. As usual, we are looking at a lot of unique and creative ways to keep the Voodoo name and 3dfx in the limelight.

FiringSquad: When will we see a need for on-board geometry processing? Are CPUs increasing in speed fast enough where on-board geometry becomes irrelevant?

3dfx believes that the most immediate problem that needs to be solved in 3d acceleration is with fill rate limitations. Geometry processing is an important technology whose time is approaching. I do not think that the real solution to the issues surrounding geometry processing has been introduced yet. People have jumped on the T & L bandwagon, but recent articles have shown the weaknesses of today's solutions, especially with the more powerful processors. There is more to geometry processing than just transform & lighting and they are independent technologies and should not be viewed as the same problem. We will be addressing it with future problems with what we feel is the correct solution.

FiringSquad: Is 3dfx planning on releasing a new revision of VSA-100 (more chips, more RAM, etc.) to keep up with the 6-8 month product cycle? Was easy scalability one of the factors in the decision to go multi-chip?

We are not announcing future parts or plans at this time. It is no secret that we have a couple of parts in the queue outside of VSA-100. 3dfx believes that we have offered a complete product line with the single, two chip and four chip configurations.

There are other ways to get life out of the VSA-100 besides adding RAM or more chips. Things like die shrinks and new features have historically been seen in the progressions of a graphics chip. Changes like that will not affect the scalability of the VSA-100, so we could still see the same benefits with any new version that could be planned.