GPU Pro 7

2

IV

Physically Based Deferred

Shading on Mobile

Ashley Vaughan Smith and Mathieu Einig

2.1 Introduction

In order for graphical applications to achieve maximum performance and therefore

maximum graphical quality, they need to utilize memory bandwidth as best as

possible. This is especially true on mobile devices without large, fast DDR RAM,

like discrete GPUs have, and where power is limited through battery life.

This bandwidth bottleneck is even more tangible in the context of deferred

shading renderers, where large G-buffers need to be stored and retrieved multiple

times during the rendering process. It is possible to take advantage of the fast

on-chip ...