I think within the next 10 years, a high-end desktop will be able to render photorealistic video in real time (say, 24 fps). My benchmark would be rendering a completely CG scene from one of today's movies - think LOTR, or some such. You certainly won't mistake it for reality, as it'll still be rendered to a screen, but you won't be able to tell the difference between it and a shot taken from a video camera. That's the direction that games like DOOM3 are taking us. Remember, the resolution requirements aren't that high. Digital video rendered for a 60 foot cinema screen is only 2048x1556, or perhaps double that for very high end work, so we aren't talking orders-of-magnitude increase in the number of pixels to generate theatre-quality video. The problems are the models and algorithms needed to mimic what we see in the real world, and the CPU cycles to run them.
/frank