LOSTCIRCUITS

Benchmarks:

Using benchmarks to show the quality of a video card is becoming rather absurd, at least in the high end graphics sector. There are other issues that are more relevant, especially those pertaining to image quality, the new hybrid technique of using "Brilinear instead of Trilinear Filtering" . With respect to other quality issues, Brent Justice and Kyle Bennett have done an excellent job on [H]ard|OCP to which we can hardly add anything new.

• 3DMark2001SE (3.30)
• 3DMark2003 (3.20)
• CodeCreatures
• Unreal Tournament 2003
• Comanche4
• Gunmetal
• Aquamark3
• X2-Rolling Demo
• FinalFantasy XI Benchmark2
• MDK2

3DMark2001SE

Instead of showing pages and pages and, er, more pages, we have taken the Result Details for the three graphics cards and stuck them together in the same screenshot, for those who are interested in numbers, this should suffice.

Default Setting: 1024 x 768 x 32bpp

Left Column: 9800XT, Center: 9800 Pro, Right: FX5900

Default Setting: 1280 x 1024 x 32bpp

Left Column: 9800XT, Center: 9800 Pro, Right: FX5900

Default Setting: 1600 x 1200 x 32bpp

Left Column: 9800XT, Center: 9800 Pro, Right: FX5900

The entries of interest are the Nature vs. Game1 "Car Chase", where the two "brands" are in opposite corners, as well as the "Advanced Pixel Shader" where the GFFX takes a rather brutal beating. Also note the design-dependent single texturing fill rate limitation of the GFFX5900. In gaming situations, it is hardly conceivable that this may matter, it is rather impossible that frame rates will show a cut-off because of single texture fill rate limitations. Keep in mind that 3DMark2001SE only tests DX8.1 features.

Next Page:    => Adding AntiAliasing / Anisotopic Filtering =>