Nvidia GeForce 6800 GS reference card
Real overclocking potential
Review The GeForce 6800 GT was a cracker. While Nvidia wowed everybody with the 6800 Ultra, it was the 6800 GT that most people bought, offering most of the bang without quite as much of the buck.
However, Nvidia has decided that the 6800 GT has had its day. The reason? Well, ATI has been fighting back to compete with the GT. Firstly, there was the Radeon X800 XL, which followed the 6800 GT's blueprint of being a scaled back version of the top-end cards. But then ATI produced the X800 GTO. Unlike the Radeon X800 GT, the GTO is a 12-pipe, six-vertex card running at 400MHz, with 256MB of 980MHz GDDR 3 across a 256-bit memory bus. Over the last couple of months this has been doing damage to sales of Nvidia's 6800 series, though the success of its 7800 series has deflected many from noticing this.
The time was ripe, therefore, for a response from Nvidia and the result is the 6800 GS. Let's get this out of the way from the first: despite having the same five vertex shaders and 256-bit memory interface, the GS has 12 pixel pipelines, four fewer than the 6800 GT. Yet Nvidia claims that the GS is faster than the GT. How is this possible? This is due to the GS possessing much greater clock speeds, with the reference card running at 425MHz core and 1GHz memory, compared to 350MHz and 1GHz for the standard GT. Clock cycles are much cheaper than pipelines and the 75MHz increase ensures that it can compete with the GT despite the reduced number of pipelines.
The reason Nvidia is able to push the cards clock speeds is due to the fact that the GS is a native PCI Express part built on a 110nm process, rather than the 130nm process of the GT, which also needed a separate AGP-to-PCI Express bridge.
Stock clock speeds haven't just increased, however. One of the benefits of the reduced production process is that there's more headroom for overclocking, and we've actually included some overclocked results to get a sense of what impact that has.
Aside from clock speeds the other benefit of the smaller process is that it's cheaper. Smaller chips means you get more on a wafer and as there's no longer any AGP-to-PCI Express bridge chip, the transistor count is reduced further. These cost savings are passed on to the board partner who passes them on to the customer, which is why the GS retail price is able to come in at less than that of the GT. Everyone's a winner!
Power draw is also reduced, with Nvidia stating that only a 300W power supply is required for a single card and 420W for SLI. Of course, as a recent part the GS enjoys all the 6-series goodness such as PureVideo support for improved video playback.