8800GT G92, 8800GTS G92, the new masters of DirectX 10

Page 5 - Some problems, Benchmarks used

Based on the software and hardware selection, we tested, as was expected a lots of problems and remarks arose. The previous page showed the initial list of video controllers, we used for comparison in the tests. MSI’s 8600GTS could not fit the MSI’s P35 NEO 2 – FR motherboard (the big, passive radiator, projecting from the board and connected to the core with several heatpipes covered 3 of the RAM slots). The reason is mainly in the fact that the memory utilized high radiators, but after all testing the altogether slow card (compared to competition) with only 1 RAM module would render all equivalency in benchmarking missing and at the end we didn’t test it at all. The same was with the 2600XT, which was expected to be corresponding to the 8600GTS in question.

On the other hand the 2900GT (unfortunately our only AMD/ATI of the Radeon 2900 series) had incredibly strange problem, for which we couldn’t find solution. The controller ran with Catalyst 8.1 drivers at 511MHz core / 1022MHz memory. The built-in overclock tool in the Control Panel of the driver displayed PASSED when testing whether the video card will run the set frequencies (default at the case) but when the button was pressed to apply these frequencies (650MHz core / 1600MHz memory), absolutely nothing happened. We also tried Catalyst 7.12, but it warned that it could not work with the current adaptor (again very strange). AMD ATI still have many problems with their x64 drivers as many of the tests on the HDs proved. At the end after several painful hours we put away the video controller in its box and the final list was:

Albatron 8800GT
MSI NX 8800GT - T2D512E - OC Edition
MSI NX 8800GTS - T2D512E - OC Edition

XFX 8800Ultra
XFX 8800GTS 640MB Alpha Dog Edition (slightly overclocked version of 8800GTS)
Sapphire HD3870 512MB GDDR4
Sapphire HD3850 512MB GDDR3
Sapphire HD3850 256MB GDDR3

For more specific problems during given test, we decided to leave a field for remarks, where all the problems are described. We also included what is relatively the best combination of settings at which the tested game plays fluently.

For performance testing we chose 4 games, which have DirectX 10 support, resulting in different effects and intense video controller load, getting the most out of them. The list could be a lot longer, but to no much gain. The main idea here was to cover engines, who use HDR (High Dynamic Range) effects, Shader Model 4.0 and other DirectX 10 features.

We tested the games at maximum detail and resolutions 1280x1024 with 4x Anti Aliasing, 1680x1050 with 4x Anti Aliasing as well as 1680x1050 without Anti Aliasing.