In the first instance, the information about a model is converted into a set of mathematical definitions that the computer and the graphics subsystem can understand. This numerical information is then manipulated in order to arrive at another set of numbers, which are the coordinates of the polygons of the model that are seen from the viewpoint of the user. In effect, this mathematical hocus pocus takes a virtual 3D world and maps it flat for rendering to a screen.

Interactive 3D creates the illusion of movement by changing the viewing plane for a scene in response to the inputs of a user. The user moving around a virtual world is redefining the flat area onto which the virtual 3D world has to be mapped. The technical term for changing the geometry that defines objects in a scene is what we call transformation. When any change in the orientation or position of the viewpoint is desired, every object in a scene must be recalculated relative to the new view. In a fast-paced game the performance is evaluated by the high frame rate. In other words, you want inputs to create instantaneous results on the screen. A transformation occurs one frame at a time, and the more frames you have the faster the overall response of the system. In the past, game developers would fluctuate on frame rates, settling for lower, but acceptable performance in order to create a greater amount of scene complexity. More scene complexity means more polygons means more number crunching.

That's simple enough. It makes sense to have hardware accelerated T&L, and take the computational load off of the CPU, which is frankly not up to the task of doing the job anyway. As a result, the CPU is left free to do all of that other good stuff that adds to the interactive experience, better AI, better physics, better audio and so forth.

So, give it up game developers. Start handing off your geometry to the graphics card.

Pow! Right there is the first problem. It's the same since the first bits of the 3D pipeline found their way into consumer graphics products. Game developers have to go off and figure out how much of their 3D geometry they're willing to give up, and then decide on how much any particular vendor's implementation of hardware accelerated T&L can handle. It's been the same at every inflexion point in the graphics business for the last five years. Every time DirectX moves up the 3D pipeline, developers become wary of handing off their valuable vertices, even though the hardware guys are saying, "Look, it's a standard. You can do it, now. Come on, guys, give us your vertices).