In case you found the last two pages somewhat confusing, here’s a graphical overview comparing the two rendering methods. Suppose we want to render a scene that consists of three 3-dimensional shapes as shown. Each shape may be composed of hundreds or even thousands of polygons.

http://media.hardwareanalysis.com/articles/small/10132.gif" alt="Sample Shapes" border=0>
The Immediate mode renderer would simply render and texture each polygon in each shape separately, then check their depths against the z-buffer in order to decide which ones are at the front and which are at the back. Once it realizes that the yellow triangle is at the front, followed by the blue circle, with the red square at the back, it will draw the scene, and simply discard the parts of the blue circle and red square that aren’t visible. Very clean and very easy, however it means that the card wasted time rendering and texturing part of the blue circle, and at least half of the red square, when they weren’t visible to begin with.

http://media.hardwareanalysis.com/articles/small/10133.gif" alt="Immediate Mode" border=0>
By contrast, the Tile-based renderer checks the objects’ depth (z-position) first, and then renders and textures only the parts of those objects that will be visible. In other words, the polygons that compose the right-hand side of the red square are never rendered or textured.

http://media.hardwareanalysis.com/articles/small/10134.gif" alt="Tile-Based" border=0>
It’s visibly more efficient, particularly in a complex scene where there are polygons on top of polygons on top of other polygons. However, as previously stated, making sure the correct polygons are drawn is no small task. For example, suppose you wanted to draw a pool of water. The Tile renderer would have to realize that the polygons that make up the water are clear, and therefore that it needs to draw the polygons behind those clear polygons (the cement foundation of the pool) as well, even though there are other polygons in front of them.