Reverend's Pulpit heeft een interview gepost met Gary Tarolli van 3dfx, waarin uiteraard weer de bekende vraag langs kwam wandelen. Wat is beter: anti-aliasing of Transforms & Lighting?
Q: I'll make an assumption here and say that the Voodoo4/Napalm won't incorporate a geometry processor. The point being it won't help much in most games and huge fill-rates is more important. Another point is that game engines need to be coded for transform & lighting in the first place. Meaning there aren't any or isn't going to be many games that will have that and therefore an on-board geometry processor would be a waste. How do you respond to people that says, "But isn't this the same as when 3dfx said, with the Voodoo3, that 32-bit rendering is not important because not many games will have that color/texture depth?".
A: It is different. First of all, 32-bit rendering slows you down but doesn't cost much more in terms of chip area, while geometry processing tends to speed things up but does cost a lot more in terms of chip area and final product cost. Before the argument went something like this - 16-bit for speed versus 32-bit for quality. Now the argument will go like this - geometry engine for more triangles vs. anti-aliasing for quality pixels. These are two VERY different tradeoffs. I foresee another giant newsgroup battle. What people don't realize is that there is no "right" answer - it's a very subjective decision as to which is better for a game player. For existing games, we feel strongly that geometry acceleration doesn't provide as much value as anti-aliasing. For future games, it's a closer call, but more geometry without anti-aliasing is only going to increase the number of aliased pixels on the screen, which is not a good thing.