Bij Reverend's 3DPulpit is een interview verschenen met Gary Tarolli van 3dfx. Gary is een van de oprichters van 3dfx, en draagt daar momenteel de indrukwekkende titel Chief Technical Officer/Chief Scientist. Dit interview is niet zoals de meeste interviews wat hersenloos geblaat over hoe goed 'we' wel niet zijn, en daarnaast is hij met 30 antwoorden inclusief vragen ook aardig uitgebreid. De onderwerpen die aan bod komen hebben vooral betrekking tot de techniek achter de videokaart en de 3D beeldkwaliteit zoals tile based rendering, (uiteraard) FSAA, TnL en de kwaliteit van reviews op het internet. Hieronder een kleine selectie van vragen uit het interview:
In a past interview you said that FSAA, Massive Multi-Texturing and Overbright lighting are features that developers are looking for. Now can you tell us what Overbright Lighting is and why developers want it so much ?
Yes, and is this still the case with developers? Other than the possibility of using multiple lightmaps, if we have T&L, why would we need massive multitexturing?
Yes. We constantly talk with developers and I am still convinced that these features will result in a better game experience than any other set of features. Overbright lighting is defined as allowing for colors to exceed the normal [0,1] range within the 3d pipeline and perhaps even the framebuffer. This means the graphics pipeline can better represent the huge dynamic range of lighting that exists in the real world.
Having T&L doesn't change the equation - T&L allows you to have more triangles in the scene, which would typically be used to add more objects. It doesn't improve the quality of the rendered pixels. Thus, T&L and the features I mentioned are really very different and independent. Even with T&L, you still need massive multitexturing and other features such as overbright lighting to improve the quality of the pixels.
3dfx has been suffering from great looking yet unappreciated babies like the 22bit post filter and FSAA. Both were (still are - Rev) unappreciated because both are something you have to see and experience to understand, 1000s of words can not explain how much of an impact both technologies have. Has this problem of marketability of these features hurt your products' success ?
Yes, I think our products' success has been hurt. I agree with you that these features are still not appreciated by as many people as we would like, and they are something you have to see in person to truly value. It's been a challenge marketing these features, but we haven't given up. The truth will prevail! We have recently added some great demos to our web site that show off FSAA.
Can you explain the basic vision behind multi-chip... can it ever be cheaper than a single chip solution?
Sure - multi-chip can be cheaper than single chip whenever a single chip gets very large. Chip yields are such that they drop off very quickly after a certain die size, so two small chips can be cheaper than one large chip. And if you are starving for memory bandwidth, two chips allows you to have more memory bandwidth without increasing the pin count per chip, which can get very expensive after a certain point. That is why you don't see single chip 256-bit memory interfaces in the consumer market.