|
/ Guides / Q3Test Deluxe Taser Edition Tweak Guide / Page 4
|
Video Options and Performance Costs
|
Ok, from here on out, unless otherwise specified, all tests were run with the following settings: 16-bit textures, texture detail at the ¾ point on the slider, geometry detail high, bilinear filtering, lightmap lighting, 640x480 resolution, low sound quality, marks on walls on, ejecting brass on, dynamic lights on, light flares on, identify target on, high quality sky on, and sync every frame off; the TNT system was run in 32-bit color depth, and the Voodoo2 system in 16-bit color depth. This configuration is considered the "baseline" against which all other changes are calculated for percentage of loss or gain in performance. The game was run after a reboot on both systems and all tests were run without exiting the game - changes to resolution, sound, or graphics settings were left to be handled by Q3Test.
Bilinear Filtering vs. Trilinear Filtering:
|
Bilinear |
Trilinear |
% Loss/Gain |
| Demo1 16-bit |
38.3 |
36.3 |
-5.23% |
| Demo2 16-bit |
52.9 |
51.7 |
-2.27% |
| Demo1 32-bit |
40.0 |
40.0 |
0% |
| Demo2 32-bit |
48.3 |
48.2 |
-0.21% |
The TNT card takes a modest performance hit for trilinear filtering, and this is more evident on the more geometrically complex Q3Test1 map. Voodoo2 cards seem to get "free" trilinear filtering. The results on demo1 are absolutely identical in both trilinear and bilinear, and the difference for demo2 is insignificant. I haven't been able to tell much of a visual difference between trilinear and bilinear on the test maps, so I recommend bilinear filtering.
Compressed, 16-bit, and 32-bit Textures:
|
16-bit |
Compressed / %L/G |
32-bit / %L/G |
| TNT Demo1 |
38.3 |
38.3 / 0% |
36.7 / -4.18% |
| TNT Demo2 |
51.9 |
52.4 / +0.96% |
51.9 / 0% |
| V2 Demo1 |
40.0 |
40.0 / 0% |
40.0 / 0% |
| V2 Demo2 |
48.3 |
48.3 / 0% |
48.2 / -0.21% |
The results of this test were surprising for me, primarily for the Voodoo2's numbers. The identical numbers could be due to superior texture management, well written texture handling in the drivers, or it could be from the 256x256 texture size limitation of the Voodoo chipset. Either way, there's no reason for either card to use compressed textures. 16-bit is probably your best choice.
Resolution: 640x480 vs. 800x600:
|
640x480 |
800x600 |
% Loss/Gain |
| TNT Demo1 |
38.3 |
24.5 |
-36.03% |
| TNT Demo2 |
52.6 |
40.3 |
-23.38% |
| V2 Demo1 |
39.9 |
28.1 |
-29.58% |
| V2 Demo2 |
48.0 |
41.7 |
-13.12% |
Concentrating on the demo1 scores here, this is a massive performance hit to move up in resolution. The Voodoo2 seems to suffer less from the increase in resolution, but the TNT is also in 32-bit color, so that might account for its nastier drop in average frame rate. TNT2 and Voodoo3 cards can probably handle 800x600 or even higher without nearly as much of a hit in performance, but those of us with standard TNT cards or Voodoo2s shouldn't even think of running higher than 640x480 unless we want to sacrifice a lot of visual quality.
|