Tile Based Rendering

The key to delivering improved graphics performance is to analyze the complete graphics 
sub-system to identify the bottlenecks that limit performance. This analysis clearly shows 
that traditional 3D accelerators are limited by their available memory bandwidth.

This situation will only get worse as new titles add complexity. More polygons will increase 
the depth complexity, advanced titles will require 3 or 4 or more texture layers and use 
better filter algorithms requiring more texture samples. The effect is an exponential increase 
in memory bandwidth requirements.

However if you step back and look at the system again, it is clear that traditional 3D 
accelerators texturing and then rendering pixels that are later thrown away and are 
never displayed on the screen. What a waste!

Textures every pixel — visible or not

Clearly a smarter approach is required to conserve and make most effective use of the 
critical resource that is memory bandwidth.

Only textures visible pixel

Tile based rendering, featured across ST's family of KYRO 3D graphic accelerators uses a 
new algorithmic approach to efficiently use this resource. Detecting which pixels are visible 
early in the 3D graphics pipeline, ST's KYRO family textures and renders only the visible pixels, 
nothing else. This efficient use of memory allows the KYRO family to deliver superior performance
from the available bandwidth.

Maximum theoretical performance on an Advanced 3D Title using 175 MHz SDR Memory