Articles :: How to Benchmark a Videocard ::

Doc Overclock · 12-01-2003 · Category: Guides

3dmark 2003

Earlier this year in February, Futuremark released the latest version of 3DMARK, 3DMARK 2003. While there has been a lot of controversy surrounding 3dmark 2003 and NVIDIA cards, it's been mainly of NVIDIA's own doing. When 3dmark 2003 was released, NVIDIA briefed every major site about what they felt was wrong about using 3dmark 2003. Had NVIDIA simply left it at that, I, as well as many reviewers would have noted the scores and left it at that.

NVIDIA, however, chose to make some questionable optimizations in the benchmark on the driver level. I won't go into the politics of it, except to say that I think the tools that 3dmark provided to identify the optimizations are important. Taking the camera off the beaten track showed missing textures in places where there shouldn't be. Why is that important? The reason for a benchmark in the first place is to have the same workload, or as near to the same workload as possible for each video card running the same benchmark. Otherwise, it wouldn't be a fair comparison, and the scores not compatible with each other. To counter this Futuremark released a patch that corrected the optimizations, patch 330. They released a new patch 340.

Installation of this benchmark was easy. Double click the 3DMARK 2003 icon and it'll install. You'll need to download the 330 patch separately and install it after the benchmark. Note, Futuremark locked out the resolutions except the default 1024x768 32 bit. If you need the other resolutions or settings, or want to change the tests being run you need to purchase the professional license.

Much like 3dmark 2001, 2003 has a set of 4 tests. The tests consist of a flight simulator type game using DirectX 7 features like single texturing and Hardware T+L, a space combat game using DirectX 8.1 features like limited pixel shaders, a Role playing like game with anisotropic hair and a Mother Nature test with more extensive use of pixel shaders including some DirectX 9 PS 2.0 ones. To run the test, simply double click the 3dmark 2003 shortcut on your desktop and click Run 3DMARK.

The score again is an amalgam of the 4 tests with differing weights for each benchmark. Taken as a whole, when the benchmarks are run on the exact same system, except for the video card, it's a useful tool. It's also been shown to be a pretty accurate indicator of DirectX 9 game performance, at least as much as can be done by a synthetic benchmark.


  1. Introduction
  2. Dronezmark
  3. SpecViewPerf 7.0
  4. Final Fantasy XI The benchmark
  5. Serious Sam: 1 & 2
  6. UT2003
  7. Quake3 Arena
  8. 3dmark 2001SE Build 330
  9. 3dmark 2003
  10. Gunmetal Benchmark 2
  11. Halo
  12. Aquamark 3
  13. Conclusion