Tile-Based Rendering: Whoopin' of Brute Force

This is where Tile-based Rendering comes in, arguably one of the best (if not the best) solution to the memory bandwidth restrictions of today's video cards. It could be argued that making the memory faster and faster (for ex. by using RDRAM) is a reasonable solution for the future video cards. But why insist on more expensive ways when the same results can be achieved by modifying the rendering technique? Tile-based rendering is established on this premise, on which it makes use of a different algorithm to processing 3D graphics. Here are the key benefits achieved by using TBR:

Eliminating excess z-buffer access: 

PowerVR KYRO makes use of "display list rendering". Instead of texturing polygons one by one as they are sent to the rendering pipelines, the polygons are grouped together in a display list, which allows the scene to be divided into small regions or tiles which have 32x16 pixel dimensions. Since these tiles are a small part of the whole scene, z-buffering can be performed using an on-chip z-buffer (24-bit). With the addition of an on-chip frame buffer for blending operations, a great deal of unnecessary external memory accesses are eliminated.