EVGA had an interview with Tim Sweeney of Epic games about the Unreal Engine 3:
Q. Which next generation graphical features do you think will make the biggest impact on gaming?
A. I feel like the evolution of gaming and graphics could almost be decoupled at this point. Today’s engines and rendering feature sets are sufficient for building just about any type of game, just as today’s movie cameras as sufficient for filming any kind of movie. This isn’t to say that we’re done! Graphics quality will continue to improve at an impressive and scary rate, and I feel that we’re still 10,000X short of having “enough” computing power for rendering 100% realistic scenes. But I think the days when “rendering feature X enables gaming innovation Y” are mostly over. This is why the use of middleware engines has increased so significantly – because one tech codebase can indeed satisfy the needs of maybe 80% of the types of games being built.
On the rendering side, I’m looking forward to the convergence of CPU and GPU programming models around next-generation uniform computing chips that are capable of both tasks. Such a convergence will enable a long string of software innovations that have thus far been held up by the Microsoft/NVidia/ATI obsession with adding new fixed-function features to the rendering pipeline. That’s my wild prediction for the next decade! I believe many of the things that were hinted at in the late 1990’s will finally start to happen – very efficient deferred software renderers, sub-pixel triangle rendering, analytic antialising, and novel new rendering schemes around voxels, frequency-space volumetric rendering, pseudo-realtime radiosity solvers, and so on.