Raytracing

The potential of raytracing is incredible. Raytracing, to my knowledge, is the best/only way to get photorealistic results out of computers. Check out this example.

So raytracing is cool and capable of quite a lot. This article goes some way toward explaining how now might be the time for raytracing.

Except it misses a big point. Raytracing on general purpose processors is incredibly inefficient. As the article itself states, the raytracer only does a single physics calculation over and over again. Other graphics technologies need to be versatile to be able to do pixel shading and texture mapping and all sorts of things, but the raytracer just does one thing over and over again. A general purpose processor's power would be wasted in operations like that, because it isn't designed for physics calculations, specifically.

Rather than simply follow the assertions of the author, I'm going to suggest that the time for raytracing is now, but that's going to have to mean that raytracing chips need to appear on the market as well. Doing raytracing on the CPU is not the way to go.