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

I don't get your last statement, it has been proven that CP-forced AF is free on ATI cards and even yeilds a performance increase with Humus's tweak

 

Humus is the man!

I first want to say thank you to Humus. It was nice of him to use his OWN personal time to help out us ATI owners. The game looks better then before and is running great.


One question though, sorry if it's been asked before but I'm getting dizzy looking though the last 10 pages i've missed...

But, can you use this tweak and still play multiplayer? Can you also host a server when having this tweak?

Thanks

 

Yes, without the Humus hack. Look on page 58 for benchmark examples. The Humus hack speeds things up because it discards the texture lookup table - ATI's CP AF does AF on all stages of it which incurs a 15-20% hit. However, using app-controlled AF like HardOCP did yields similar performance results as the Humus hack.

Anyway, I just did my own benchmarks, Nvidia cards gain about 10% with app controlled AF:

p4 3.2ghz stock
1gb ram
6800GT stock

1600x1200 8xAF High Quality Forced AF: 51.7fps
1600x1200 8xAF High Quality App-Controlled AF: 56.7fps

 

Re: Humus is the man!

It was reported that the first tweak had problems... no one has said anything about the latest tweak. I guess you'll have to try it to find out

 

Well, Carmack himself has stated that even minor changes will knock nV3/4x "off the fast track" because nVidia's drivers are heavily optimized for the D3 engine. I think it's the old "out of sight, out of mind" phenomenon. People see Humus making visible changes to the D3 config text files (which in this case are nothing but common WinRar-compressed text files), but they don't see the heavy optimizations buried deep in the nVidia drivers, completely out of users' control, so they mislabel Humus' actions as "shader replacement" (when all he's done is change the D3 config files as will certainly happen with all D3 mod creators to one degree or another), while being blissfully unmindful of the nVx driver optimizations relative to D3 over which no one but nVidia has any control whatever (not to mention knowledge of.)

What's interesting to me here is that since even Humus' minor config-file changes reportedly adversely affect nV4x D3 performance to a measurable degree, while at the same time positively affect R3/4x0 performance to a measurable degree, it really isn't clear at all at this point whether the case is that nVidia's drivers are heavily optimized for the D3 engine, or whether the D3 engine is itself heavily optimized for the nVx drivers, or whether it's a good bit of both... It's fascinating to unravel, though.

 

Here you go ruined.


pxc posted

a drop of 10fps useing the tweak . Obviously it is breaking shader replacements .

 

No one is doubting that the driver (or more correctly, the hardware) is capable of proper AF. What is in doubt is if the "Application Preference" CP mode actually functions as it should. Several people have suggested that it doesn't with recent drivers.

 

So then ATI would gain quite a bit of ground on Nvidia in the Hardocp benchmarks if they had of used Control Panel AF (which looks much better) and Humus's tweak.

 

How so ?

control panel af provides a higher image quality than doom3 controled af.

With the tweak its faster than doom3 controled af .

I don't see why it isn't functioning right

 

Obviously, it can't possibly the technical explanation offered by several B3D members.

So jvd, now that we've explored that, back to the issue you've tried to sidetrack.

ATI gained no significant performance improvement from the Humus' hack when using App controlled AF, hence there will be no significant improvement or gap-closing in the HardOCP benchmarks. Hence if ATI fixes their drivers, the Humus hack will cease to be all that useful.

 

Aside from yourself, Ruined, and Triniboy; who exactly has suggested that?

 

Yes those are two big ifs though. One big question is "Does x800 forced AF look better than 6800 app AF in Doom3?"

 

Please stop the bullshit. I never suggested that. Just because I didn't gang up with WaltC and jvd to bash in Ruined's head doesn't mean you can lump me in with anything he says. Thanks!

 

Dig, if you are suggesting that ATI would gain some large amount of performance from the reduced quality seen in Doom3 app AF, that has been proved wrong by benchmarks in this thread which show that with the Humus Hack, whether it is app-controlled or CP-forced, the performance is the same; but the quality is still crappier with app-controlled AF.

 

Or the tweak is simply slower on NV40. If NVidia's FX composer outputs the correct numbers, It won't get faster than the original shader (but several changes are equally fast).

 

We haven't explored anything .

What are the reasons they say ?

I'm a beyond3d member and I say its driver replacements .


Btw before humus tweak 16x aniso 2x fsaa 61.1fps (all forced in control panel) after tweak 65.8 fps with them forced again.

x800xt pe

 

hey um... this may or may not be of interest...mind you I don't have a benchmark, but for Max Payne 2, if I set the AF to app pref, and set AF ingame, it *seems* to run less smoothly than if I had set AF in the CP @ 16xAF....

I just used the framerate counter in game and looked at one scene and... yeah......so I could be wrong..

anyhoo, the nvidia shader replacements certainly couldn't have been introduced in say.... a 44.xx driver would they? Why not test the earlier drivers on the the NV3X to determine 1)when the shader replacements were introduced and 2) how bad the default shaders are. Obviously this investigation could not be done on the 6x.xx drivers... unless by fluke the release drivers didn't include the shader replacements.

 

Again...the "problem" is that I don't have any other games which suggest to me that this might be the case, and so no basis on which to consider it as a plausible possibility. Hopefully, you aren't suggesting that the Catalyst team somehow managed to have App Preference misfire on D3 alone of all the games in the universe?... Just seems to be no empirical basis for that supposition.

 

See the post above yours and the other posts he's made.

Wow, I'm a beyond3d member and I say its not, guess that means I'm right too eh? Logic plays no part in reality, just whether or not you are a beyond3d member.

Try using app controlled AF instead of forcing it without the Humus tweak. That is where you will see a similar boost.

 

I was actually asking Steve to back up what he said about recent drivers, and I'm implying that there is a problem with the way D3 calls for application AF with ATi rather than it being a driver problem since AF seems to work pretty damned peachy when applied thru the control panel.