Introduction

It has been a long time since Nvidia and ATI released new architectures so close to one another. The last time this happened was in 2004, with the R420 and NV40 launches, and overall the two architectures were extremely similar (six vertex shaders, 16 pixel pipelines, 16 texture units, etc.) Since that time the two companies have followed divergent paths with their architectures, but while they may not have necessarily agreed on technical choices, they both stayed with the concept of monolithic GPUs – with each new generation, the number of transistors have more or less doubled, and then, based on these enormous chips, a full line of cards was produced by varying the number of processing units.

No doubt encouraged by the success of its G80, Nvidia has chosen to continue using that approach as you know if you’ve read our article on the GT200. ATI, on the other hand, since it was bought out by AMD, has had a number of difficulties, in particular with its R600 architecture, which didn’t perform as expected and caused many technical difficulties for the engineers. With its financial problems, it was hard for AMD to continue battling it out with Nvidia, whose sterling financial health enabled them to continue developing such chips. So instead of continuing to focus on raw performance, AMD decided to concentrate on two factors – performance per watt and performance per mm² of die. Have they succeeded?