Conclusions + Analysis

This content was originally featured on Amdmb.com and has been converted to PC Perspective’s website. Some color changes and flaws may appear.

Conclusions

What can we conclude based on what we have seen here today.

  1. NVIDIA’s claim of a performance increase with v44.03 in Unreal Tournament 2003 is true.
  2. The performance increase stated on the NVIDIA website applies only to Unreal Tournament 2003 and not with Freelancer or Morrowind.
  3. Using v43.45 drivers, both GeForce Ti4800SE and FX5600 execute anisotropy functions properly (i.e. increase in anisotropic sampling increases texture sharpness).
  4. Using v44.03 drivers on the GeForce Ti4800SE, anisotropic filtering performs properly in Freelancer, Morrowind, and Unreal Tournament 2003.
  5. Using v44.03 drivers on the FX5600, anisotropic filtering performs properly in Freelancer, and Morrowind.
  6. Using v44.03 drivers on the FX5600, anisotropic filtering DOES NOT function properly in Unreal Tournament.

Based on the 6 conclusions above, we can therefore state that we have accomplished our stated objectives.

  • Determine if the optimization is isolated to Unreal Tournament 2003.
    Yes, because Freelancer and Morrowind are unaffected.

  • Determine if the optimization is isolated to the FX5600 card.
    Yes, because the Ti4800SE is unaffected.

  • Determine if the optimization is isolated to v44.03 drivers.
    Yes, because v43.45 drivers are unaffected.

  • Determine if optimization’s effect on image quality.
    Yes, anisotropic filtering is not consistently applied across Unreal Tournament 2003 (causing textures to remain unsharpened).

Analysis

Having reviewed and worked with graphics hardware as both a moderator at AMDMB and as a hardware evaluator, it would appear to me that there is a correlation between the stated performance increase and the degradation of anisotropy functionality in the FX5600 using v44.03 (“There’s no such thing as a free lunch”). The fact that when using v43.45 drivers anisotropy works properly (i.e. all textures are sharpened), enforces this correlation.

Why did this happen? I can think of a few reasons:

  1. There’s a bug in the driver that causes anisotropic filtering to be broken. The correlating performance increase is just a coincidence.
  2. The particular combination of Unreal Tournament 2003 and v44.03 drivers causes the problems. Otherwise we would see if functioning properly like we see in Morrowind and Freelancer.
  3. NVIDIA intentionally programmed the driver this way.

What ever the reason is, I think it is necessary that consumers, analysts, and manufacturers be aware that the current v44.03 driver, though increases performance of UT2K3, has a corresponding problem with anisotropic filtering that fails to function properly.

Where do we go from here?

The issue investigated today does not end here. Since we only tested the FX5600, it may be possible that the FX5900 and FX5200 are unaffected by the anisotropic anomaly (but I doubt that).

I encourage other websites, consumers, and interested parties to particpate in this matter by conducting their own tests of the v44.03 drivers. If companies release buggy or inferior drivers, it’s our responsibility as consumers and the press to let the community know (as it may affect their purchasing decisions). AMDMB.com will gladly assist/consult individuals wishing to do their own experiments.

I welcome NVIDIA to comment on our findings. I sincerely hope the issue is a bug in the driver and can be easily fixed with a subsequent revision.

As always, I invite any comments or questions you may have on the topic by participating in our video card forum.

Thank you for reading AMDMB.com.

Editor’s Note: Just to keep everything in check here, I’d like to say that this information was given to NVIDIA several days before this publication and we are awaiting their response to it. Whether this is a bug or not remains to be seen, but we thought users should have the all the information they can before choosing which version of graphics drivers to install. As the HardOCP article that was published earlier states, the differences you’ll see while actually playing the game may be slim to none, but we feel that any kind of frame rates should be compared on as equal an image quality as possible — and with the two varying results from the two Detonator versions shown here, that wasn’t the case.

« PreviousNext »