Nvidia apparently cheating 3DMark with drivers

Discussion in 'Video Cards & Monitors' started by glimmerman, Jun 24, 2008.

  1. glimmerman

    glimmerman Member

    Joined:
    Nov 3, 2005
    Messages:
    2,203
    Location:
    Perth
    I suppose ATI does the same sometimes as well, but I don't think it's been on this kinda scale.

    Link
     
  2. Falls

    Falls Member

    Joined:
    Jun 28, 2001
    Messages:
    3,287
    Location:
    Gun Cat @ Bikini Atoll
    Storm in a tea cup.

    Nvidia have Physx support on high end GPU's which gives a much higher score in the Physx type test in 3dmark vantage.

    The writer in the article seems think that nvidia needs to jump through Futuremark hoops to release drivers.

    It seems to me 3dmark vantage testing methodogies and implemtation sux.

    F.
     
  3. silverbullet

    silverbullet Member

    Joined:
    Apr 2, 2005
    Messages:
    2,002
    Location:
    Marcoola, Sunshine Coast
    OK before i begin i could'nt give a toss about nvidia or ati either way :weirdo:


    Now, it's not the first time Nvidia have errrm cheated with drivers they have been pwn'd before over it.

    As far as FutureMark is concerned, say if your benching your system using said driver and then you go to "register" your benchmark results would'nt it get rejected if your using unapproved drivers?
     
  4. TaroT

    TaroT Member

    Joined:
    Jan 18, 2002
    Messages:
    8,705
    Location:
    Hazelbrook nsw 2779
    no it will let you register it
    i use beta all the time and no issue
    and yes 3dtoss who cares....average joe/joeline punter cares

    they are where the money is coming from
    they look at the marketing hype
    3dtoss 3 billion times faster....and they buy it.

    As said before it is unethical.
    all people need to knwo is how it performs in games.
    especially heavy physics game slike ut3 and graw and cellfactor etc etc.
    does it give that big ooomph or not.
     
  5. Kaine[zof]

    Kaine[zof] Member

    Joined:
    Mar 22, 2004
    Messages:
    2,176
    Location:
    Melbourne
    ... and people wonder why i don't give a flying fuck about 3dtoss/futurewank/etc benchmark results and always wait for actual in-game figures.
     
  6. itsmydamnation

    itsmydamnation Member

    Joined:
    Apr 30, 2003
    Messages:
    10,141
    Location:
    Canberra
    i see the big issue as being that the test no longer reflect reality for single card setups. because you have 100% gpu resouces dedicated to the phyX calcs in cpu test 2 it scores very high but its not doing any graphics work.

    What will be interesting is to compare phyx vs GF in vantage and then in phyx games.
     
  7. MR CHILLED

    MR CHILLED D'oh!

    Joined:
    Jan 2, 2002
    Messages:
    136,455
    Location:
    Omicron Persei 8
    No, it's not simply 3dmark, it's sales and that is money. And probably significant amounts if you can tear the opposition a new one.

    nVidia being unethical? No never :Paranoid:
     
  8. HyDrA

    HyDrA Member

    Joined:
    Jul 1, 2001
    Messages:
    3,821
    Location:
    SA
    Thats all I had to read to dismiss the article.

    That dude is an ATI/AMD fanboi...
     
  9. kini_mini

    kini_mini Member

    Joined:
    Sep 28, 2004
    Messages:
    1,812
    Do people still pay attention to 3dmark results though? I would have guessed that the sort of person who took them seriously wouldn't be the sort of person who bought and installed their own cards, or even read reviews in the first place. Why wouldn't you skip to the real-world performance sections? :confused:

    Anyway, I seem to remember that many moons ago ATI were caught out doing something similar with the Radeon 7500 I think it was...I think that was the first time and they've been taking turns ever since :p
     
  10. Murmandamus

    Murmandamus Member

    Joined:
    Jun 27, 2001
    Messages:
    948
    Location:
    Sydney
    I don't see this as cheating. If a card/drivers do something that provides an advantage in overall system performance without compromising image quality then it should be reflected in the results.

    With nvidia running PhysxX on the GPU the situation has changed. 3dmark needs to change to keep up if they want to have any relevance, not the other way around.
     
  11. Saiyan_Overlord

    Saiyan_Overlord Member

    Joined:
    Jul 6, 2005
    Messages:
    1,999
    Location:
    Sydney 2074
    I don't really see how better drivers is cheating and how nvidia wanting to implement physics processing on their gfx cards is cheating either.. How is it cheating to include such features? If we say nvidia is cheating to optimize their drivers for 3Dmark, we might as well say that they are cheating in every game or 3D app by optimizing their drivers... Rather pointless argument as we all enjoy newer drivers that bring fixes + performance boosts :tongue:.
     
  12. MR CHILLED

    MR CHILLED D'oh!

    Joined:
    Jan 2, 2002
    Messages:
    136,455
    Location:
    Omicron Persei 8
    What really defines cheating in the 3dmark context anyways? Surely if there is a problem with the software then futuremark need to address it.
     
  13. Smakked

    Smakked Member

    Joined:
    Nov 28, 2007
    Messages:
    2,071
    Location:
    Goldcoast
    Sounds like an ATI fanboi article I bet Charles wouldn't post if ATI were cheating. Its a cut throat market all business cheat against each other just alot you never hear about.
     
  14. itsmydamnation

    itsmydamnation Member

    Joined:
    Apr 30, 2003
    Messages:
    10,141
    Location:
    Canberra
    it's very simple really. you have a fix amount of resources on the gpu, now you do test 1 where it uses 100% of the shaders to do graphics and produces score X, then you do test 2 where it uses 100% of the shaders to do math and produces score Y. Using these two score to determine your final score isn't a fair comparison because your GPU doesn't have 200% resources.

    the only way i see it as being legit is to have a gpu just for phyX ( not in sli etc)
     
  15. TOFUGil

    TOFUGil Member

    Joined:
    Jun 30, 2004
    Messages:
    1,998
    Location:
    somewhere
    yer, fanboi alarms went off, were told this guy has it in for Nvidia

    Ditto, its not something sneaky, Futuremark will just need to change how their scores are tally up.
     
  16. r-type

    r-type (Banned or Deleted)

    Joined:
    Sep 1, 2006
    Messages:
    616
    Like Nvidia needs FutureMark's approval for anything. :lol:
     
    Last edited: Jun 24, 2008
  17. power

    power Member

    Joined:
    Apr 20, 2002
    Messages:
    56,697
    Location:
    brisbane
    i'm really impressed with what nvidia are doing with their drivers atm - awesome stuff.

    oh and OP it's allegedly not apparently.
     
  18. Saiyan_Overlord

    Saiyan_Overlord Member

    Joined:
    Jul 6, 2005
    Messages:
    1,999
    Location:
    Sydney 2074
    The GPU itself though is 100% resources... There's no argument to say that the test can't be designed to do both physics + 3D work at that exact same time... Intel must be cheating by buying out havok if that's the case :o. Gee, all this so called "cheating", simply because people do not appreciate the fact that the company is adding new features to their products and adding more to already existing products. People who would otherise call this cheating are kidding themselves because they have an issues that the competing product was not considered or does not have such features :thumbdown:. I also might add that AMD must be cheating by having DX10.1, something that nvidia doesn't LOL![/sarcastic]

    I still fail to see why it's cheating for nvidia to implement physics acceleration even despite the fact that it was still in development to run via CUDA by the time 3Dwanker Vantage came out. It's just like having the argument "ATI has DX10.1 but nvidia has physics acceleration" and in AMD's case, DX10.1 worked in their favor before AC got patched.
     
  19. banshee

    banshee Member

    Joined:
    Sep 8, 2001
    Messages:
    1,695
    Location:
    Springwood, West of Hell (Sydney, actually)
    What Itsmydamnation said. :p The only way to get a realistic score for this would be to run the PhysX test with a graphics load on the GPU as well.

    Fair enough that nVidia have set the drivers to run the physics calcs on the GPU, as they own it all now, but the results from the way Vantage work fall to their advantage because of 3dMardks 1 subsystem at a time approach. It needs a whole system test to be more applicable to actual usage.
     
    Last edited: Jun 24, 2008
  20. itsmydamnation

    itsmydamnation Member

    Joined:
    Apr 30, 2003
    Messages:
    10,141
    Location:
    Canberra
    hummmmm, your so far off the mark its not funny. the simple fact is the test is scored in a way that gives a massive advantage to a phyx gpu based system when in reality it isn't there because it doesn't have the resouces to give to it.

    please note i never said nvidia cheated, but up until now vantage has aligned very well with peformace in games. now it doesn't. If DX 10.1 was giving the same results because it had tests just for it then it would be in the exact same boat.
     

Share This Page