When you view a 3D scene or play a 3D game on your PC, all you actually see is a 2-dimensional image. The image appears to be 3D, but the monitor is only able to display a 2D image because of the fact that it is flat. For many years, programs were designed, modeled and drawn in 2D. They looked much like a painting or a board game, everything looked flat against the screen.

A few years ago, programmers and game designers realized that although the monitor can only display 2D images, they can still be designed to give the appearance of 3 dimensions to give the images more life. The problem with this was that it took a lot of power from the computer. Even though the image on the monitor was only 2 dimensional, the computer still had to render the whole image in 3D. This meant that everything, even objects covered by other objects in front of them, had to be rendered. Following this, the Video Card had to decide which parts can be seen and which can't. For example, if a box was covering a part of the image behind it, the Video Card would have to recognize that. Also other parts, such as the back of objects, would have to be invisible in the final image sent to the monitor.

Thanks to unperfected depth perception by the Video Card, a player in this online multiplayer game seems to be on top of a rifle, three times bigger than him, which another player is holding.
Screenshot taken from the game "Counter-Strike"

There is an analogy that can be used to explain what the Video Card does to make a 3D image. This analogy is taking a picture of a car. First, the whole car has to be built from scratch. Next, the picture is taken from the front. Although the picture is 2D, it seems 3D because it was taken of a real car. To create this 3D-realism time had to be spent to make the car from scratch. The same thing would have to happen inside the Video Card. It would have to build a 3D object, and then it would only use a fraction of the whole object in the picture.

The hardest problem is to overcome the depth perception. A 2D scene has height and width, while a 3D image also has depth. The computer has to decide which object should be visible. This is determined on the object's depth, distance away from the camera, as well as which objects are blocking any parts of it. If flawless precision is not maintained, people would see problems with the images, such as seeing through walls, or objects not having the right perspective.

Since this kind of precision takes a lot of power, and the CPU could not handle it alone, a processor similar to the CPU, but specialized to handle 3D image rendering, was developed. It is basically used to render everything, and then decide what is and what is not seen.

This Video-Card features the most powerful 3D Accelerator GPU, the Geforce3

Immediate Mode Rendering is today's industry standard for 3D rendering of images. Using the example with the car, we can see that this is a very inefficient way to render images. In the typical Video Card, the whole car is built as a wire-frame model. Then, the whole car has textures applied to it. This is the real problem, because it is the most time-wasting thing the Video Card has to do. Following the application of textures to the car model, the Video Card needs to decide what is seen. The parts which are seen are sent to the monitor, while the rest of the data is useless. Immediate Mode Rendering exploits the memory speed problem further by doing things the way it does. Since 2/3 of the 3D image are not seen, the Video Card uses only uses 1/3 of the data it processes.

This example shows you that to render the top image, all three shapes on the bottom have to be created and textured. Is there any more efficient approach to 3D rendering? Slowly, a new approach is becoming more widely accepted. This approach is Tile-Based Rendering.