How do 3D Cards Work?

On the most simple level, your PC's processor receives information about the image the software you're running is trying to show. It then sends that information to your graphics card, which judges how best to display it, and in turn sends its instructions to your monitor. If that sounds simple, it's not. Sorry.

The major way of rendering a 3D scene depends on a system known as ‘rasterization’. First, a wire frame of everything in the scene is generated. Then those wire frames are converted into triangles. A frame consisting of a ton of triangles will make for a more complex shape, but less triangles means less work for the graphics card. Next, every pixel associated with each triangle is filled in - rasterized - to create the appearance of solid objects. Then textures -image files provided by the game to plaster over surfaces - are applied to the triangles, and distorted and their overlaps blended as necessary. Then pixel shader effects kick in, further altering each pixel's appearance. The net result is a 2D image containing instructions for the monitor about exactly what color each and every pixel should be. And that's just one frame - for smooth gaming this needs to happen sixty times every second.

A graphics processing unit with a higher clock speed can run this whole gauntlet a little more quickly, while shader pipelines can process per-pixel and vertex effects in parallel, but as game scenes grow ever more complex, that increased rendering speed is offset by the card having to do more hard work.

To ensure said smoothness, the card actually gets ahead of itself, rendering a frame or two ahead and storing them in its frame buffers - its memory. This memory is also used for storing large texture files from the game; newer games seeking the increased detail that higher resolution textures offer is one of the reasons we're seeing quantities of 3D card RAM slowly increase.