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

Ummmm jvd, I would think that after 40 pages you would realize that nobody is debating the worth of Humus' change when you are forcing AF. This discussion should have been over 20 pages ago.

 

but ATI gains in quality when using this tweak don't they ? I'm referring to the bilinear / trilinear discussion a couple of pages back

 

Mabye.


Sadly though ruined only focuses on putting ati in a bad light and he refuses to dive deeper into the nv40 performance issues.

IT could be that its simply slower. It could be that its simply slower becasue it breaks shader replacements.

 

Yeh, the point is that when using *app controlled AF* ie setting the 16xAF in doomconfig.cfg *not in the ATI CP*, results in a similar performance boost to using the Humus tweak.

 

just in case some of you missed this, and I think it's something worth looking at...well... in my humble opinion :

 

Your argument is funny. How do you know that NVIDIA's driver isn't already changing the shader to something close to what Humus' is doing? Then there wouldn't be much performance difference because they are already doing the same work. If you're willing to concede that NVIDIA is doing shader replacements for the 5900s, then why wouldn't NVIDIA create a hand-tuned shader for the 6800s while they were at it?

-FUDie

 

Delve into the performance issues of the Nvidia card that currently has the best D3 performance in a thread about an ATI performance hack

 

And you do vice versa

 

The point is hardocp used app controlled. with humus's changes you can run cp controlled af with better image quality than app controlled.

By doing this you will get performance gains and image quality gains .

Which should be benchmarked against nvidia's and which ever way nvidia's af looks better.



Some of you just don't like the fact that ati has some gains happening and image quality boosts . Ruined has been out to ruin this tweaks good works .

By spewing crap out of his mouth.

 

The 6800 has no need for hand-tuned shaders, that's why. I'm not sure what Nvidia could even benefit from by hand tuning shaders on Doom3 when the game was designed on Nvidia hardware. The 5900 is at a SEVERE deficit with complex shaders and needs to have hand tuned shaders.

 

Isn't it bad enough Carmack stated that the drivers were optimising things that could break if the engine changed?

But I agree--this should be investigated; even if id claimed those optimisations were valid, it would be interesting to see what NV felt worth changing for performance's sake.

93,
-Sascha.rb

 

Who says ?

 

Put down the pipe.... Step away from the pipe...

 

I think it's fair to say that I'm just repeating what other people have said -- I can't make any outright claims myself as I don't have ATI hardware at my disposal.

I'd be a little surprised if both IHVs aren't bending the rules, but WaltC seems only willing to believe that NVidia are -- I'm trying (in vain) to persuade him that that may not be the case.

Edit: typos

 

Correct.

Incorrect, at least nothing significant (more than 1-2 fps)

Nvidia looks the same with app controlled or CP-forced AF.

No, I'm just posting reality. Humus tweak is good for right now since its the only way to get full AF quality with ATI cards at decent performance. However, comparatively, since HardOCP used application-controlled AF for ATI cards, Humus' tweak will offer little to no gain from the results in those benchmarks, hence there was really no "closing the gap" as dig suggested earlier. Instead, Humus' tweak is just a helpful tweak for ATI users to get better IQ, but overall ATI's performance has not changed from the HardOCP benchmarks. HardOCP used 4.9 beta + app controlled AF, that will give the same performance (fps) results as 4.9 beta + humus tweak with or without app controlled AF. Those are the facts.

 

you hit it right on the nose

 

except there are gains . For me it was 3-4 fps between cp af and almost 6 with control panel af vs game af.

 

Again, we aren't talking about CP AF, as HardOCP used appAF, and that is what gives similar performance results to Humus' Tweak.

 

Your not listening .

I get performance increases even against app af .


They are not 40%

but they are there . Mostly 2-3fps with forced af vs doom3 af.

 

Ok, 2-3fps, that is more realistic. Do you really think 2-3fps is closing the 10-20fps gap that is seen between NV and ATI cards in Doom3?