SE: How would you best describe what how the T-Buffer is implemented?

Scott Sellers: In very general terms, the implementation of the T-Buffer in hardware is that we now have the capability of managing many frame buffers in a unified memory architecture. Since we can render into multiple buffers, we can now render multiple view-points or time samples, so we can generate depth of field, motion blur, etc. But it's the magic of how these multiple buffers get recombined back together before being output to the monitor that is what makes the T-Buffer effects so new and exciting.

When performing spatial anti-aliasing, the multiple buffers that the T-Buffer architecture manages are allocated to be used for rendering at the sub-pixel level. As a result of being able to render at the sub-pixel level, we can remove the jaggies so common in today's 3D Accelerators.

SE: Can these (and the other effects enabled by T-Buffer) be used at the same time or are they mutually exclusive?

Scott Sellers: All the effects that we discussed can be implemented simultaneously. None are mutually exclusive…

SE: Are these really 'cheats' to approximate the effect of the real thing with as little overhead as possible? If so, what steps will be taken to educate developers on how to take full advantage of them without running into their inherent shortcomings?

…depends on what you mean as "cheats." Obviously everything we do in 3D hardware acceleration is coming up with ways to mimic real-world phenomena. The T-Buffer is a way to mimic real-world cameras and the human eye, which allows us to do the special effects we've described.

The spatial anti-aliasing capability of the T-Buffer does not need evangelism to the game development community. The user will have the ability via a control panel setting to force anti-aliasing for a given game. The other effects, such as motion blur and depth of field, do require evangelism to the development community. We actually feel this is the ultimate "1-2 punch" because users get the immediate out-of-box experience of being able to immediately upgrade their installed base of games with the anti-aliasing capability of the T-Buffer, but we also add value to that consumer's purchase of the hardware further down the road when games start to ship with support for the advanced effects like motion blur and depth of field…

Remember that no one educates developers like 3dfx. We've had to do it since the inception of the company. We introduced the world to real-time, perspective correct texture mapping in 1996 with Voodoo Graphics, and then were the first to market with multi-texturing in 1997 with Voodoo2. We've always been able to introduce new architectures and evangelize to developers to use those new features - the T-Buffer is no exception, and we fully expect tremendous support from the developers in using the advanced effects.