Leaked Nvidia Drivers 44.67 doubles 3DMARK2003 Score

SpinnerSpinner Birmingham, UK
edited July 2003 in Science & Tech
Apparently the recently leaked Nivida detonator drivers 44.67 seemed to have directly addressed performance issues under Futuremarks benchmarking software 3DMARK2003.
We ran the GeForceFX 5900 through 3DMark03, first with the current 44.03 driver, and then with this new 44.67 driver. The performance increases seen in 3DMark03 were substantial, almost double on one of the tests. We also tested three of the games we usually use for 3D testing. We found essentially no performance difference between the two builds there.
www.extremetech.com

The full report:
http://www.extremetech.com/article2/0,3973,1188119,00.asp?kc=ETRSS02129TX1K0000532

Original Short-Media news post with download links for the Detonator 44.67 drivers:
http://www.short-media.com/forum/showthread.php?s=&threadid=1254

Comments

  • ThraxThrax 🐌 Austin, TX Icrontian
    edited July 2003
    Leaked doesn't mean they haven't been altered by a third party in between.

    It's an outlandish statement, but considering all the **** that's being slung, nVidia is not dumb enough to do the same thing yet again.
  • mmonninmmonnin Centreville, VA
    edited July 2003
    Sounds like some special 'optimizations' are in place.
  • SpinnerSpinner Birmingham, UK
    edited July 2003
    These benchmark performance increases must be just effecting the Geforce FX line of cards, understandably of course, because my 3DMARK2003 scores on my GF4Ti4400, weren't that much different from previous driver versions. Special optimisations indeed.
  • WuGgaRoOWuGgaRoO Not in the shower Icrontian
    edited July 2003
    i still think ati is better :)
  • EnverexEnverex Worcester, UK Icrontian
    edited July 2003
    If the quality isnt degraded in the benchmarks then essentially its not cheating, but if no other games are effected then it sort of makes 3DMark even more useless than it already is (for comparing real world performace).

    NS
  • SpinnerSpinner Birmingham, UK
    edited July 2003
    NightShade737 said
    If the quality isnt degraded in the benchmarks then essentially its not cheating, but if no other games are effected then it sort of makes 3DMark even more useless than it already is (for comparing real world performace).

    NS

    I think 3DMARK2003 is yes, at the moment a rather poor benchmark in terms of getting an idea of real world performance, like you said. However, that will change I think when games as standard evolve to contain the many graphics utilisations and advancements which 3DMARK2003 tests. '2003 is still just a little ahead of its time, that though I am sure, will change.
  • Geeky1Geeky1 University of the Pacific (Stockton, CA, USA)
    edited July 2003
    I don't run 3dmark03 anymore. I run 2001se instead. Why? Because after a while, a 3dmark score of <3000 gets depressing. :D
  • GHoosdumGHoosdum Icrontian
    edited July 2003
    Regardless of the fact that '03 is ahead of its time, if nVidia is once again making drivers purely to raise benchmark scores, that's cheating.
  • ThraxThrax 🐌 Austin, TX Icrontian
    edited July 2003
    OMFG.

    LEAKED RELEASE.

    Everyone chill out.
  • SpinnerSpinner Birmingham, UK
    edited July 2003
    GHoosdum said
    Regardless of the fact that '03 is ahead of its time, if nVidia is once again making drivers purely to raise benchmark scores, that's cheating.

    I'm not sure that that is the case here. Nvidia have stated previously that their cards 3DMARK2003 performance doesn't to justice to it's real world performance, in comparison to other cards performance ratios in the same benchmark. The increase in scores in this case, may just be below par scores being elevated to what they actually should be to offer a realistic comparison for real world gaming.

    However, as people have previously pointed out. 3DMARK2003 is yet to establish itself as a real world benchmark. So essentially part of what I have just said becomes slightly moot.
Sign In or Register to comment.