For the initial set of benchmarks, I just wanted to compare frame rates between the two systems, and to see how a TNT in 32-bit color measures up against a Voodoo2 in 16-bit color. I've been playing Q3Test ever since I got my Celeron in 32-bit color, simply because I like eye candy. For me, a big part of Quake III is for the game to look good. If I want ugly, I can play Quake 2, or Barbie Fashion Designer.

I started with the playing configurations we had both been using prior to these tests. For mine, I was using the High Quality settings in Q3Test, with two changes - 640x480 resolution, and 16-bit textures - I ran the tests in both 16 and 32-bit color. For Dakota's system, it is in (of course) 16-bit color, with compressed textures, bilinear filtering, texture detail at the middle-point of the slider, ejecting brass turned off, and light flares turned off. All tests for this entire guide were run 3 times each and the results recorded and averaged.

	TNT	Voodoo2
Demo1 16-bit	46.8	48.2
Demo2 16-bit	55.8	49.7
Demo1 32-bit	38.4
Demo2 32-bit	50.1

Dakota's Voodoo2	Joe's TNT

This didn't surprise me too much. My system was generally slower as I have more of the visual options enabled. I didn't think this was really fair on a card vs. card comparison (especially since my system more or less lost this showdown). I decided that for things to be fair, the benchmarks should be run with identical settings. First, I set my system to the same settings as Dakota's - compressed textures, bilinear filtering, texture detail at the middle-point of the slider, ejecting brass turned off, and light flares turned off. Here's how that stacked up:

	TNT	Voodoo2
Demo1 16-bit	54.1	48.2
Demo2 16-bit	57.7	49.7
Demo1 32-bit	47.5
Demo2 32-bit	54.5

Wow. My TNT kicked ass. The TNT was faster in 16-bit color, and ALMOST as fast as the V2 (in 16-bit color) was in 32-bit color. I decided that I ought to give the Voodoo2 a chance with the same video settings I normally play at as well, minus a 32-bit color test for the Voodoo2 card since it doesn't support 32-bit rendering. Here's how things ended up with Dakota's system running with the same settings I was using initially (with 16-bit textures, ejecting brass on, light maps on, trilinear filtering, texture detail at the ¾ point of the slider):

	TNT	Voodoo2
Demo1 16-bit	46.8	39.9
Demo2 16-bit	55.8	48.3
Demo1 32-bit	38.3
Demo2 32-bit	50.1

Again, my TNT rocked the Voodoo2. It's interesting to note, however, that the Voodoo2's performance dropped much, much less on q3testdemo2.dm3. I chose to test both demos to get data on two very different game types. Since demo2 was recorded in spectator mode, it is MUCH less intensive than demo1, and we should be more concerned with the results from demo1, as it is more likely to reflect your actual gameplay.