Tile-based rendering (TBR) is 'the word' on the street, and the new found popularity is a courtesy of the memory bandwidth limitation of DDR memory we have seen in video cards such as the Geforce2. Tile based rendering is nothing new though, for over three years now Imagination technologies has been implementing it in their graphics chips and since then they have been hollering about it's miraculous memory bandwidth saving solutions. But two or three years ago gamers couldn't be bothered, and for very good reasons too. We were happy with how our old cards rendered games, and memory bandwidth limitations weren't a huge deal. More importantly, the video cards that offered the technology didn't really do it for the average gamer in terms of performance. The most recent TBR board we saw previously, neon250 (PowerVR series 2), didn't have much more to offer other than some buggy drivers and sub par performance, and TBR certainly didn't help it much. Although it is noteworthy that while the neon250 wasn't a big success on the PC graphics market, it currently powers Sega's DreamCast which has sold millions, and we have seen some very visually impressive games on the console. We could start ranting about how tile-based renderers can do this and that, but the true appreciation for TBR is only realized when compared to the average Joe rendering style of the other video cards. So a clarification of how a 3D accelerator works in general is essential.

The Basics: The Conventional 3D Graphics Processing - A virtual 3D world, such as in 3D games, needs to be mathematically processed into a 2D image that is displayed on our monitors through different stages. In a basic sense, the application or platform provides a map or blue print if you will, of this 3D world. The 3D processor understands this lay out as vertices that have a designated coordinate in 3D space. Objects in this 3D space are modeled by polygons that are defined by their surrounding vertices. Now depending on the observer's view of the 3D scene, the 3D data has to transformed for each frame. Next stage is lighting, which simply processes the light sources in the scene and how these lights affect the appearance of the scene. Subsequently comes clipping that removes the vertices that are not part of the observer's field of view. Note that clipping does not mean removing the areas that are going to be behind other objects.