Introduction

The mainstream shootout that never was

By all accounts, NVIDIA’s GeForce 6600 GT has ruled the mainstream roost ever since it was first introduced last summer. The GeForce 6600 GT merged high clock speeds (500MHz core/500MHz memory in PCI Express variants) with an eight pipeline architecture and shader model 3.0 support, all in a package that retailed for $200. The end result was a product that established a new benchmark in mainstream performance, pumping frames faster than a high-end RADEON 9800 XT in some cases; and to top it all off, PCI Express 6600 GT cards could be combined together for SLI for nearly double the performance.

ATI’s answer to the GeForce 6600 GT was the RADEON X700 XT. The RADEON X700 XT matched up well to the GeForce 6600 GT, featuring similar clock speeds and a somewhat similar architecture. ATI did a few things different than NVIDIA with X700 XT, going beyond the GeForce 6600 GT’s design in many ways, but ultimately ATI just couldn’t produce the X700 XT in the volumes needed for a mainstream part at the lofty clock speeds they had set and the project was canned, with the few X700 XT chips that were produced eventually finding their way into higher end X700 PRO cards like the Sapphire X700 PRO TOXIC we reviewed two weeks ago.

PowerColor X800 (bottom) and MSI GeForce 6800 PCI-E

The PowerColor board poses with X700 PRO, GeForce 6800, and GeForce 6600 GT

To ATI enthusiasts, the X700 XT was yet another poster child for all that was wrong with the company. Just as X800 XT/X800 XT PE were announced with much fanfare only to ship to retail in very limited quantities, the X700 XT was another embarrassment for ATI.

But little did the public know that ATI had a trick up their sleeve in the form of the RADEON X800!

The RADEON X800 core

First announced last December, the RADEON X800 borrows most of its design cues from its older brothers in the RADEON X800 PRO/XT, only it has been designed from the outset to be cheaper to produce. You’ve got the same 2.0b shaders, with 3Dc support and a 256-bit memory interface (four 64-bit memory controllers to be exact), only the RADEON X800’s R430 core is based on TSMC’s 0.11-micron manufacturing process.

PowerColor X800 and GeForce 6800 GT

PowerColor X800 and X800 XL boards

The smaller process allows TSMC to produce more chips per silicon wafer (assuming similar yields), thus reducing manufacturing costs. The only real downside to 0.11-micron for end users who are enthusiasts and like to overclock is that the process isn’t tuned for high clock speeds like the older 0.13-micron process was, which featured performance enhancements like low-k dielectric. As a result, the graphics core on the X800 and X800 XL is 400MHz. However, the beauty is that the cost savings introduced by moving down to 0.11-micron allows ATI to cram 16 pixel pipelines into the X800 XL, while the X800 features 12 pixel pipes (one of its pixel pipeline quads is disabled).