Best utility for testing 3D performance?

The actual in-game FPS counters are the best way to test the "speed" of your 3D card. Winbench is not that great of a program. As mentioned, several card companies have been known to optimize their drivers for benchmarks. Case in point, STB just recently did this to their TNT drivers for the Velocity 4400 so it would be faster. A lot of drivers now are WHQL (Windows Hardware Quality Lab) certified. Basically the drivers are tested for "tweaking" and compatibility problems. Another example of Winbench 98 problems is that I remember reading a review of the Starfighter 3D card, it beat out a single Voodoo2 card in overall score. The marketing dept then used the winbench number (whatever it represents, I don't know) in their adverti***ts. Funny thing is, Voodoo2 is faster in every game I have seen (e.g. Quake2, Unreal, Forsaken, etc.). I would stick to frame rate counters in games for the test of speed. That is the only true test of a card and how it performs against others. Tom

>> Which benchmark software do you use to test your PC's 3D performance? Is >> Final Reality still the best? > I use 3D Winbench 98. You can download the big beast, but what I did was >go to the Ziff-Davis site and order the CD for something like six bucks >(shipping included.) It really is a good test, I don't know if any of the >current cards will pass _all_ the tests Problem with 3D Winbench is that it is so well known by the manufacturers that they (well, many) aren't above "tweaking" their drivers to achieve better performance on the test. Meanwhile performing no better, and in some documentated cases *worse* than previous drivers did, in actual software. Most gamers have learned to ignore 3D Winbench and instead favor benchmarks found in games. Quake and Quake2 have become standards due to the popularity of the games and the availability of the demos. Being the simulators group, I'll mention that GPL (rendition and 3dfx/Glide support) has a framerate counter as does CPR (D3D). Flight sims with framecounters include FS98 (D3D), Jane's F15 (3dfx) and I believe F22 ADF.

Quoted text (10 lines)

> Another example of Winbench 98 problems is that I remember reading a review > of the Starfighter 3D card, it beat out a single Voodoo2 card in overall > score. The marketing dept then used the winbench number (whatever it > represents, I don't know) in their adverti***ts. Funny thing is, Voodoo2 > is faster in every game I have seen (e.g. Quake2, Unreal, Forsaken, etc.). > I would stick to frame rate counters in games for the test of speed. That is > the only true test of a card and how it performs against others. > Tom

And then PCComputing reviewed the Monster3D2 (Voodoo2) 12MB card and reported that it was slower than the Intel i740-based Starfighter. In real game benchmarks the Monster will perform from double to up to quadruple the i740. Of course, Intel also owns part of Ziff-Davis publishing (PCComputing's publisher) but I'm sure that had nothing to do with it...