Raytracing Explained in 40 minutes

Last time we talked about Rasterization (the process of drawing triangles to the screen), today we talk about Raytracing, which is the other major approach to doing 3D graphics.

Rasterization has the downside of having a cost proportional to the number of triangles drawn to the frame buffer (even if they don't appear, they can get overdrawn). Raytracing by contrast has a running time proportional to the logarithm of the number of triangles, which is much much faster, and can do actually realistic lighting, reflections and refractions, but these three things add to the rendering cost of a scene.

The general notion is that you send out one ray (or more) per pixel and when you hit something (this is accelerated by an RTX card) you can shoot out additional rays to do things like see if the surface can see a light, or if it has a reflection from another object, or if it is translucent send an internal ray bending through the object, and so forth. The cost can go up exponentially with the number of bounces you allow, so real time raytracing makes quality tradeoffs versus speed in terms of how many rays it will use.