How to boost your Doom3 performance by 40% on ATI hardware

Discussion in 'PC Gaming' started by Humus, Aug 8, 2004.

  1. ATIpwnz

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    14
    Likes Received:
    0
    Can someone plz post the file with demirug's change or something? I'm so lost... Are u supposed to delete all of Hubus's changes and copy and paste his stuff? And if so where? Can someone copy paste it?
     
  2. tEd

    tEd Casual Member
    Veteran

    Joined:
    Feb 6, 2002
    Messages:
    2,098
    Likes Received:
    64
    Location:
    switzerland
    like this
     
  3. BlueSteel

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    7
    Likes Received:
    0
    This one is faster:

    Code:
    # perform a dependent table read for the specular falloff 
    # TEX R1, specular, texture[6], 2D; 
    SUB_SAT R1.x, specular.x, 0.75; 
    MUL     R1.x, R1.x, R1.x; 
    MUL_SAT R1.x, R1.x, 16.0; 
    
    # modulate by the constant specular factor 
    MUL   R1, R1.x, program.env[1];
    Notice the ".x" behind the parameters. Not sure what this does (unfortunately I still don't understand shadercode :cry:) but it makes the whole thing a lot faster.

    -BlueSteel
     
  4. advman

    Newcomer

    Joined:
    Apr 29, 2003
    Messages:
    11
    Likes Received:
    0
    No, as was just shown, at least on 6800GT there is no difference between APP/CP, whereas on ATI without the tweak (which is actually a fix as I see it now) you have the choice between abysmal performance and severe texture aliasing :roll:

    I guess I have to go to sleep now. Anyway, it would be helpful if someone could repeat the experiment on Cat 4.7. I wonder if this is the cure introduced with the hotfix :?

    Actually I'm a bit angry now. ATI always excuses poor driver defaults with "... but if the application requests highest quality then it gets highest quality". Now we have one of the first applications that seems to make extensive use of that and what we get is this mess :evil:
     
  5. pxc

    pxc
    Newcomer

    Joined:
    Feb 8, 2002
    Messages:
    76
    Likes Received:
    0
    Because it does affect nvidia performance and it's not an "improvement."

    At 1600x1200 HQ in timedemo demo1 (no driver/cfg tweaks) I lose over 15% (65.6fps -> 55.6fps) using the first Humus mod and lose 13% (65.6fps -> 57.1fps) with the final Humus mod. At lower resoltions where I'm CPU bound I gain a maximum of 2% (85.0fps -> 86.7fps) in 640x480 LQ. No thanks.

    Question: if the ATI driver team traces the problem to texture thrashing as Humus suggested and found a workaround, would you guys still use this mod even if the real fix was a bit slower?
     
  6. acrh2

    Newcomer

    Joined:
    Aug 8, 2004
    Messages:
    16
    Likes Received:
    0
  7. BlueSteel

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    7
    Likes Received:
    0
    I disagree. ATi's "Quality"-AF via the CP is the highest possible quality for ATi cards; which is quite visible if you enable it: all the aliasing is gone. This has also been proven by 3DCenter a while back; ATi always had some optimization for it's D3D-AF ("brilinear filtering"), but the OpenGL-AF was always as good as it can get (on ATi-hardware that is...).

    So it seems the problem is the way Doom 3 "requests" AF from the driver, obviously it doesn't request it in the same way the CP does. And it IS a little bit faster than the quality-option in the CP, it's just slower than the performance-option (which still looks way better imo).

    -BlueSteel
     
  8. BlueSteel

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    7
    Likes Received:
    0
    Try this:

    Code:
    timedemo shot_demo001
    
    timedemo shot_demo002
    Renders just one frame, and always the same.

    -BlueSteel
     
  9. grecco_julio

    Newcomer

    Joined:
    May 23, 2004
    Messages:
    233
    Likes Received:
    0

    What's your video card ? :shock:
     
  10. Recall

    Newcomer

    Joined:
    Jul 6, 2004
    Messages:
    89
    Likes Received:
    0
    I think it has been determined that this tweak does not have any performance increase on R3X hardware or below. We have seen 6800 lose no IQ between CP and Game AF. Could someone with a X800 series card now take the same shot please?
     
  11. Blito

    Newcomer

    Joined:
    Apr 30, 2004
    Messages:
    12
    Likes Received:
    0
    I beg to differ. Many people with 9800s and 9600s reported gains
     
  12. logen999

    Newcomer

    Joined:
    Jul 24, 2004
    Messages:
    57
    Likes Received:
    0
    Location:
    Queens, New York
    Rev said something about posting his received e-mails from JC about this subject. I don’t see them anywhere, are they posted?
     
  13. Solr_Flare

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    9
    Likes Received:
    0
    I got a performance boost here as well on my 9800xt. It was a minor boost, but a boost all the same of about 4fps. Not as huge a boost in the AF department, but still all around a nice pickup at no dicernable loss in image quality.
     
  14. TheAlSpark

    TheAlSpark Moderator
    Moderator Legend

    Joined:
    Feb 29, 2004
    Messages:
    21,577
    Likes Received:
    7,126
    Location:
    ಠ_ಠ

    ....and what happened to those UE3 questions we all posted in that thread that Dave started.... :?:


    ..........
     
  15. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,573
    Likes Received:
    657
    Location:
    WI, USA
    I think the 9600's main problem is memory bandwidth. Especially on my notebook. The fog and smoke just murders it. Fillrate at 445Mhz isn't exactly overly tiny....

    I DEFINITELY see higher framerate spikes with Humus' tweak.

    However, the super plasma/rocket glow tweak does get a little out of hand when the plasma gun has 15 rounds in the air. Goodbye framerate! I shoot in pulses instead, aim better as a result, and save ammo! :)
     
  16. acrh2

    Newcomer

    Joined:
    Aug 8, 2004
    Messages:
    16
    Likes Received:
    0
    I've done testing on this system:
    p4 3.0 @ 3.5 GHz, 2G ram, x800 xt pe @stock.
    Catalyst 4.7, "high quality" in D3

    timedemo shot_demo001
    normal, APP in control panel - 41 fps , artifacts
    normal, 8x AF in control panel - 36 fps , no artifacts
    humus, APP in control panel - 43 fps , artifacts
    humus, 8x AF in control panel - 43 fps , no artifacts

    timedemo shot_demo002
    normal, APP in control panel - 49 fps , artifacts
    normal, 8x AF in control panel - 43 fps , no artifacts
    humus, APP in control panel - 51 fps , artifacts
    humus, 8x AF in control panel - 52 fps , no artifacts

    Effect of APP vs forced 8x AF in normal - forced AF is slower by 12%, but the filtering artifacts are removed in forced.

    Effect of Humus's on IQ - highlights are just a tad brighter, which actually makes the game looks nicer than original w/ its dull defaults highlights.

    Effect of Humus's on framerate - raised framerate by 5% in both APP and forced AF.

    Conclusion : Humus is brilliant. His tweak allows using CP forced AF w/ actual 5% fps boost, while eliminating artifacts and making the game look even better than original.

    I have a 7.9M zip with all the png images in labeled subfolders, but I need to post it somewhere so anyone can see this for themselves.
     
  17. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,573
    Likes Received:
    657
    Location:
    WI, USA
    The RV350 DEFINITELY sees gains in maximum framerate (without fog/smoke around) and in AF. I can run AF now. It doesn't cause a 50% performance loss at 8X, now it's more like 25% or so.

    RV3x0 and R3x0 absolutely do see gains.
     
  18. Reverend

    Banned

    Joined:
    Jan 31, 2002
    Messages:
    3,266
    Likes Received:
    24
    Most have been posted, scattered in various threads in various forums as they may be. The last email I received from JC hasn't been posted and probably never will be coz I intend to use that in an interview that also gathers all the emails from JC (on various topics regarding Doom3).

    "Interview" should be finished in a few. Up to Dave to determine when to publish it. Server is getting hammered due to this thread so dave may choose to wait a couple of days, if/when things slow down a little.
     
  19. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,573
    Likes Received:
    657
    Location:
    WI, USA
    Aww, push that new dual Xeon till it breaks!!!!!!
     
  20. Blito

    Newcomer

    Joined:
    Apr 30, 2004
    Messages:
    12
    Likes Received:
    0
    *cross fingers for a 9600xt @ 350 memory*
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...