Physics of computer games and PhysX

I miss the good old days when I had time for playing computer games for long hours. I remember that time when I first saw a book on game physics. I had never thought of them as a physics problem. It seems that the market in this area have grown since then. Now I see that the problem can be very interesting. Find an algorithm for classical mechanics so that you can allow low precision in position and momentum but still the long term behavior 'looks' reasonable and the complexity is optimized. What is the best compression (or in physics language: coarse-graining) algorithm for (classical) physical reality? There might be something deep lying there, but I assume people are just using the enormous power of today's computers. Do you remember playing less than 1MB games for days on a 286? Programming must be much more fun and challenge at that era.

That was the software part. On the hardware side, specialized processors for celestial mechanics simulations are not new for physicists. Now there is a commercial physics processor available for game physics calculations: PhysX. Do you think that a physics processor will be a standard component of computers like math and graphics chips? I believe it is an essential component for computers that interact with its physical environment. Any way, this chip is a nice event for the year of physics.