Yeah, fundamentally VR (or at least the simplest form of VR which involves only one interleved or two physical screens normal to the user's field of vision) is really not that different from any 3d game or really any other 3d application. The only real difference is that it uses two POVs, one for each eye. I'd point out that motion tracking isn't even necessary for VR (though extremely useful) and that most of the systems that we learned what VR was had roughly the computing power of the old-school Atari. The positional lag has always been due to the fidelity of the motion sensors and the functional limitations of the graphics processors at the time. Sensor fidelity has improved considerably and is even better with high-end equipment, and graphics, well, we're on a game board. The capabilities of graphics processing has obviously increased. Of course there are more advanced VR setups (the CAVE system is pretty cool, though only practical for a limited subset of applications) but the concept is still the same. I'm not sure massively multiprocessor is necessary at all, though I could see the advantage of, say, dual graphics processors, one for each eye, and in independant CPU to handle the actual mechanics and positional data.
