batracio
Junior Member
Offline
An evil petting zoo?
Posts: 69
Spain
Gender: 
|
Running that benchmark is pointless. The result will be inconclusive and cannot be compared to anything else.
Quake II uses current desktop color quality for its own rendering color depth. That is, if you are using a 16 bpp desktop resolution, Quake II will run on 16 bpp. If you are using a 32 bpp desktop resolution, Quake II will run on 32 bpp. This option can't be selected anywhere inside the game.
There are other options that can be modified within the game, and will also affect the benchmark result: screen size (it is not the same as screen resolution), texture quality, 8-bit textures, and maybe more. Both sound effects and music should be disabled before running any benchmark.
Timerefresh is a very inaccurate benchmark per se. It usually finishes in such a short time that any event in the background will spoil the result. It also depends a lot on the map place where you run it. Large, open areas with many visible items will give low numbers, while small, closed areas will greatly improve those numbers. You must run it exactly at the same place to get comparable results. There are other Quake II benchmarks much better suited for unbiased comparison, like demo1, demo2, massive1 and crusher.
Another inconsistency: your friend's GF2MX gave him 1066 FPS at 640x480 resolution, that is, 1066 (frames/sec) * 640 * 480 (pixels/frame) = 327475200 pixels/sec. GF2MX's theoretical pixel fillrate is 175 MHz * 2 pipelines (each one with 2 TMUs) = 350000000 pixels/sec. This benchmark allows your friend's card to almost hit its theoretical maximum performance, so actually he is not CPU limited at all. But your Voodoo5 got 777 FPS * 640 * 480 = 238694400 pixels/sec, while its theoretical fillrate is 166 MHz * 2 pipelines (actually 4, but each one with a single TMU) = 333000000 pixels/sec. Far from hitting its theoretical limit. You have a bottleneck, he doesn't. Apples and oranges.
If you are not interested in a fair benchmark, and just want to beat your friend's score, launch Quake II from a 16 bpp desktop, disable sound and music, locate a hidden corner in a small area, grab a grenade and run timerefresh. The secret room where you find the grenades for the first time is a perfect place. And if you don't mind to cheat, overclock the Voodoo5, enable all the aggressive optimizations, replace Quake II's 3dfxgl.dll with a faster driver, and tweak some console variables (cl_gun=0, cl_lights=0, gl_ztrick=1, gl_picmip>0, and so on). Your friend seems to be already using cl_gun=0, by the way. I can't see his laser gun. Or, for an instant win, look down at the floor and run timerefresh. Yes, it's that easy. Timerefresh sucks.
For a much more interesting GF2MX vs Voodoo5 match, you'd better run demo1 or demo2 at 1024x768x32bpp. This benchmark will stress both fillrate and bandwidth, and the poor GF2MX will bite the dust against the Voodoo5.
|
Home
