Starting with the 3.2GHz machine we can compare the results using the synthetic tests of 3DMark 2001SE and 3DMark03. In the former, itâ€™s a closely matched contest until ATi nudges ahead when we hit the higher screen resolution with FSAA fully implemented. However in 3DMark03, the 6800 GT takes the lead across the board. The difference, is that we run 3DMark03 with AF enabled as well, and under the heavier load, the four extra pipelines on the 6800 GT immediately make their presence felt. Of course we wanted to see whether this would be reflected in the benchmark scores of real world games.
The Aquamark scores seemed to mirror the 3DMark findings with nVidia winning at all tests except for being one frame per second slower at the top end test.
Tomb Raider on the other hand, belonged to ATi â€“ though this is traditionally an ATi friendly test. Moving to Halo and Far Cry, both recent and demanding games, saw nVidia come out in front again. Highlights include the 73.9fps achieved running Halo at 1,280 x 1,024 compared to 58.9 on the ATi X800 Pro.
Looking at Far Cry we noticed some anomalies compared to our ATi scores. The X800 cards were clearly not running with FSAA and AF enabled at 1,024 x 768 and 1,280 x 1,024. The ATi tests were run using version 1.0 of Far Cry, which suffers from visual corruption on the new graphics platforms - this removed textures and thus reduced the amount of work the cards needed to do. We encountered the same effect initially with the 6800 GT, but once we applied the Far Cry 1.1 patch the visual corruption disappeared, and the scores came into line. However itâ€™s telling that at 1,024 x 768 in 4x FSAA and 4x AF the 6800 GT is still faster than the X800 Pro, even though the latter card was clearly not running with these effects enabled.
Looking to the 3.4GHz scores, we have to bear in mind the differences between the two systems. Even with that considered, we can still draw some interesting conclusions. Even in a slower system the 6800GT pulled ahead in most of the tests, particularly with X2 and Halo. The 3.4GHz systems also show up that even these new super high end cards have their limitations. In Far Cry, once FSAA and AF are enabled both cards are no faster in the 3.4GHz systems than they are in the 3.2GHz PC, indicating that the cards are a bottleneck rather than the CPU.
Judging from our results, if you like to run games with image enhancing features such as anti-aliasing and anisotropic filtering, nVidia has a winner on its hands. At the same time though, itâ€™s clear that the 12-pipeline X800 Pro isnâ€™t the card that ATi would have chosen to compare against the 6800 GT. If nVidia can get the 6800 GT out of the door quickly then it is likely to force ATi to drop the price of the X800 Pro, which is good news for gamers seeing as itâ€™s still an impressive card. We can therefore expect ATi to release a lower clocked version of its 16-pipeline X800 XT Platinum Edition to match the 6800 GT.
As things stand at the moment though cards based on the nVidia 6800 GT should shape up to be the best choice for those who are looking for high-end performance but still want to keep to some sort of budget.
Weâ€™ll hold off on scoring the GeForce 6800 GT until we test a retail board. With many manufacturers already announcing 6800 GT products, it shouldnâ€™t be too long before we get our hands on a full retail sample.