Applications of Computing in Industry : Lecture

Imagination Technologies is at the forefront of 3D graphics innovation for mobile and embedded application, providing the graphics IP beating at the heart of the world's most popular and loved mobile computing devices.

Following on from last year's talk on some of the innovations we've had to make to scale down a desktop graphics architecture -- one with tens or even hundreds of watts of power draw, big die sizes and dedicated external memories -- into your pocket, we'd like to talk about the challenges of building on that to put a fully programmable teraflop of compute ability in that same space.

Graphics trends, especially in games, have dictated that one of the most valuable resources in modern graphics is the ability to spend a lot of time on per-vertex and per-pixel computation. Application of advanced lighting models, on-GPU animation systems, deferred rendering, complex post-process effects and more all demand more than a trivial amount of computation. Being able to offer the graphics programmer hundreds of flops per vertex or pixel lets them give artists the power and flexibility they need to make every piece of geometry and every resulting pixel look as perfect as possible.

It's not just a matter of waiting for Moore; real innovation in the low-level graphics architecture has to be made to achieve teraflop levels of performance, especially in a tiny mobile chip in your phone that can't sip more than a couple of watts at peak. This talk covers the challenge of getting to that point.