Virtually all graphics cards capable of processing 3D scenes have been based on a method of rendering known as Immediate Mode Rendering. It’s quite simple actually -- the system first renders and textures the entire scene in all three dimensions, and then performs a depth test to determine which of those pixels are actually visible to the camera, and displays only those pixels.

http://media.hardwareanalysis.com/articles/small/10130.gif" alt="Immediate Mode Diagram" border=0>
Fig 1. Immediate Mode Rendering

Let’s assume you wanted to take a picture of something: a house, for example. If you were to obtain that picture using Immediate Mode Rendering, you would first build the entire house, and once that’s done, you’d go out front and take a picture. It’s very simple, very logical, and will give you a flawless picture. Although your picture is only printed on a flat piece of film, it will appear much more realistic and lifelike than a painting of a house, because it was taken from a real, 3-dimensional object.

However, as I’m sure you realize, that’s a somewhat inefficient way of doing things. You had to build an entire house, but the picture only displays the front of the house. In essence, all the time spent building the back of the house was wasted, because it’s not visible from the point at which your picture was taken anyway. This is known as overdraw, when a pixel is rendered and textured, only to be thrown away because there’s another in front of it. Estimates from software designers indicate that for each pixel actually visible to the user, an Immediate mode renderer has an overdraw of anywhere from one to three pixels. That is, for every one visible pixel, one to three (we’ll say two, on average) have been thrown away because they’re not visible. When you consider that 2/3 of the work done by an Immediate mode renderer is essentially wasted, the GeForce 2 GTS’ 800 Mpixels/second fill-rate doesn’t seem quite as impressive.