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

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

  1. Solr_Flare

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    9
    Likes Received:
    0
    Well, here is the thing, up until Humus's work, there was no garanteed way to boost your performance without sacraficing something or garanteeing that you actually get a performance boost.

    Tweaking the config file with the cache stuff does not work for everyone, it seems to be hardware dependant. Uncompressing the pak files may boost your in game performance some, but it increases load times, and if your system is fast enough, the performance gain is negligable.

    The remaining config tweaks simply give up some effects or image quality in order to gain FPS.

    The work humus has done, in its current incarnation, appears to sacrafice nothing(or so little it barely matters) but gains you a performance boost of some sort no matter what.
     
  2. RockShocker

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    10
    Likes Received:
    0
    Location:
    Brantford, Ontario
    I could be wrong but I thought I read somewhere, either here or at Rage3D that Humus has not altered his autoexec.cfg or doomconfig.cfg files at all.
     
  3. daerBaer

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    12
    Likes Received:
    0
    my 9800Se @ Pro begs to differ. Had a huge performance boost at 1024*768 with 0AA, 16AF, Vsync=Off, basicly ultra settings except for image_usePrecompressedTextures "1"
    before:33.3
    after:42.2
    both values were measured after the second run of the timedemo
     
  4. RockShocker

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    10
    Likes Received:
    0
    Location:
    Brantford, Ontario
    Wow, nice increase :D Do you visually notice a difference between Ultra and High Details?
     
  5. Demirug

    Veteran

    Joined:
    Dec 8, 2002
    Messages:
    1,326
    Likes Received:
    69
    I can not do this because I did not have a DOOM III Version by my own at the moment.

    I got the shaders (as a 3da log file) because I am doing some work for a german print magazin. It was only to check how many and witch type of shaders D3 use.

    The proposed amendments which I made were only the result of general experiences. Therefore I never promised or predicted any performance improments.
     
  6. dizietsma

    Banned

    Joined:
    Mar 1, 2004
    Messages:
    1,172
    Likes Received:
    13
    While I still await my copy of the game in Elizebethan England I am rather puzzled over the range of % improvements people are seeing. For a one line code change 1.5FPS is very good, but some people are seeing double figure increases. Is it down to where you test or how you test ?
     
  7. Neverender

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    6
    Likes Received:
    0
    Now, are you saying he's hit the nail on the head, or that he's a tool? ;P

    Back on topic...

    Unfortunately, my gains with this particular tweak haven't been so stellar. My benchmarks, as I haven't bothered really doing any until now, are baffling me as well.

    B-B-ATI 9600 Pro, Barton 2500+, 512M Corsair XMS, no overclocking. I timedemo 40 FPS at 640x480/medium, 32 FPS at 800x600/medium, 29 FPS at 800x600/high, and provided I change my image_cacheMegs (which I still have a sinking feeling is a placebo effect), 800x600/high is as playable as anything else. But that scaling doesn't make any sense to me, and it doesn't seem like my poor 2500+ would be choking something as relatively ancient as a 9600 Pro.

    Changes "post-Humus" (zing) are within margin of error for me (half an FPS) with every single timedemo run, and unlike most, I can't even say I FEEL a difference in normal gameplay.

    Can someone give me an opinion on whether or not these results make sense to them?
     
  8. Ruined

    Regular

    Joined:
    Sep 20, 2003
    Messages:
    317
    Likes Received:
    4
    From what I've read it looks like less of an "incompatibility" and more of a design choice on Carmack's part to write generic shader code for all hardware instead of lots of paths for different vendors. i.e. he eliminated paths only if they were completely necessary, even dropping the NV30 path despite the increases that probably could be made with that.

    ATI in this case might benefit from an X800 path, but I wonder if that conflicts with Carmack's ideals for the engine?
     
  9. Van

    Van
    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    12
    Likes Received:
    0
    Im getting some little white dots on character models with the tweak file posted on page 15.

    I have 2.4 ghz P4, Radeon 9800XT 256 mg., 1024 mg. RDRAM, no overclocking at all. Ive got AGP set to 4x. Everything was fine before I put this tweak in.

    Does anyone know how to fix this problem besides just removing the tweak?
     
  10. Demirug

    Veteran

    Joined:
    Dec 8, 2002
    Messages:
    1,326
    Likes Received:
    69
    Reverend, i don't care about how big is the improvement if there is any at all. If it help someone to get a better gameing experience I will be happy. As a developer I know how complicated it is to find the perfect shader for each chip. Sometime even the same chip behave behaves different if you use different AA/AF settings. If your systemdesign force you to use one for all you have to find a good compromise. Because of this I prefer to build my shaders with a dynamic system.
     
  11. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    9,045
    Likes Received:
    1,119
    Location:
    WI, USA
    I actually am beginning to see this code change by Humus as more of a user mod to the game. Who ever said that a mod had to be levels, or changed weapons, or whatever. Why not performance tweaks to the engine for specific hardware?

    It is only happening because ID made the game so stunningly adjustable and moddable. In fact, the only other time I can remember tweaks happening to a game engine like this was with Quake 3 and the special AMD/Pentium DLL compilations that showed up.

    The only reason I can see JC dropping paths for specific hardware is to simplify his life and guide the industry. He went with the ARB2 path to make a standardized path that would work well for the most hardware. It makes a good target for IHVs to optimize/design for. ATI and NVidia can worry about making their OpenGL drivers work well with ARB2, instead of proprietary extensions/tweaks.
     
  12. Thowllly

    Regular

    Joined:
    Feb 6, 2002
    Messages:
    551
    Likes Received:
    4
    Location:
    Norway
    Tested my code some more, the MUL doesn't need a _SAT (of course) and the numbers 3.0 and and -2.0 worked better. Now I think it looks very close to the original, even closer than Humus. But since there is an aditional instruction it's not so fast, I only got a 2% increase over the original. If someone else would test it, that would be nice...
    copy'n'paste this:

    MAD_SAT R1,specular, 4.0, -3.0;
    MUL R1, R1, R1;


    edit:...and the instructions were the wrong way around, and I think the new numbers are better...
     
  13. Van

    Van
    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    12
    Likes Received:
    0
    Can anyone help me out with the character artifacts?
     
  14. BetrayerX

    Newcomer

    Joined:
    Aug 13, 2003
    Messages:
    74
    Likes Received:
    0
    Location:
    PR
    Pictures, plz.
     
  15. CitizenC

    Newcomer

    Joined:
    May 16, 2003
    Messages:
    11
    Likes Received:
    0
    Looks to be about the increase i'm getting with the Humus tweak on my X800PRO + 4.9B's.

    Not sure how these other guys are getting upwards of 10fps...hmmm
    I dare say a few people are being less than honest :wink:
     
  16. madshi

    Regular

    Joined:
    Jul 26, 2002
    Messages:
    359
    Likes Received:
    0
    Have you tried with 8x/16xAF?
     
  17. daerBaer

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    12
    Likes Received:
    0
    may be it's all about the anisotropic filtering :p

    I had done some tds before trying to find the optimal configuration on my 9800se @ pro and it seemed that AF was one of the major things that slowed the game down...

    I'd hold up a bible and swear by the lord ( if I hadn't burned it and danced around the fire for a while ). But honestly my framecounts aren't made up!
     
  18. Melly

    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    3
    Likes Received:
    0
    Ok...sorry for my poor English,but i do my best.

    I really want to see that light-shader of the plasma also.
    So i have unzipted the pak000.pk4 file...
    Then removed the pak000.pk4 file...
    Openend the 'def" file and then openend the "weapon_plasmagun.def" file.
    Serached for those lines and removed the " // " before the lines.
    Saved the whole thing and.......nothing happend :(
    Can anyone explain to me step by step what i have to do do make it work?
    Thx for the efford.
     
  19. Van

    Van
    Newcomer

    Joined:
    Aug 9, 2004
    Messages:
    12
    Likes Received:
    0
    Ok, I don't know how to explain this but the second time I started the game up, the white artifacts were gone. Weird.
     
  20. CitizenC

    Newcomer

    Joined:
    May 16, 2003
    Messages:
    11
    Likes Received:
    0
    Yes, Multiple resolutions, AF & AA combo's.
     
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...