3D Accelerator Review Step One - 3D Performance, the Real Deal

GLQuake BIGASS1 Benchmark - Cyrix/IBM 6x86MX PR200

Playing GLQuake on a system with 6x86MX PR 200 is quite a commitment to make. The CPU does not have enough power to supply the powerful Obsidian card with enough data so that it even falls back behind the normal (but overclocked to 60 MHz) Voodoo cards. The performance is clearly limited by the CPU power. If you really want to play GLQuake with a low end system, don't go for anything else than either Voodoo, Voodoo Rush or Rendition Verite 2x00, anything else pretty much sucks, I gotta say it this way. The worst experience was benchmarking RIVA PCI cards with bigass1.

Please notice the almost identical result of the overclocked and the non-overclocked Thriller 3D. The low FPU geometry setup power of the 6x86MX keeps the Thriller 3D from showing a better 3D performance, just as the Obsidian. The overclocking of the Stealth to 62.6 MHz still generates some 3D performance increase, but more than that doesn't improve it anymore.

Image Quality - GLQuake BIGASS1 Benchmark

Again I have to say that the only real thing is Voodoo, but Rendition Verite 2x00 cards look almost as good. Permedia 2 cards offer a pretty decent image quality, although the Dynamite 3D/GL shows some strange rasterization. NVidia's alpha 2 release of the OpenGL engine is not as dark as before anymore, but it still isn't up to compete with Verite or Voodoo yet. The PowerVR does not support the lighting in Quake, which makes it look really dull.

Conclusion - GLQuake BIGASS1 Benchmark

If you're a real Quake freak and if you've got the cash, go for the Quantum3D Obsidian 100SB to get Voodoo2 performance now. This card rocks big time as long as if you've got at least a K6 or Pentium MMX 200. Otherwise my doctor recommends Voodoo. If you want a cheap 2D/3D solution that still does pretty well, go and get the Diamond Stealth II S220 and overclock it to 62.6 MHz. This will make it as fast as an unoverclocked Hercules Thriller3D which costs double, but which has got its temptations too. It's faster and offers a lot more features.

Quake II Timedemo Demo1 Benchmark

Unfortunately there isn't any 'bigass1' for Quake 2 yet, so I had to go for the implemented demo1. However the results are pretty close to the bigass1 results in GLQuake, so that it seems as if the implemented benchmark is better than demo1-3 in Quake. It's actually the next proof for the high authenticy of 'bigass1' against 'demo1-3' in Quake.

To run the benchmark in Quake II you have to type a little bit different commands than what you are used from Quake. First type 'timedemo 1' to switch on the benchmarking mode, then type 'map demox.dm2', where 'x' stands for the number of the demo you want to run. I ran 'demo1'.

The benchmark was ran on a high end Pentium II 300 system with Abit's LX6 motherboard and also on a low end 6x86MX PR200 system with FIC PA-2012 motherboard. Both systems were equipped with 64 MB SDRAM, Quake ran without any tweaks, no autocfg file, no status bar, 640x480.