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

Does anyone know how to make two identical screenshots from the timedemo? (so I can see the exact differences in IQ in a specific frame with and without the 'Humus fix')

Btw, I got a 10fps speedboost in the timedemo with this fix on my X800XT, thnx Humus!

 

Ok, third try and yet again no white artifacts. Yay.

 

File environment.vfp in /glprogs has strings:

# this should be better on future hardware, but current drivers make it slower
#MUL result.color.xyz, R0, fragment.color;
MUL result.color, R0, fragment.color;

Have anyone tried to test it? What were the results?

Can I consider R800 as "future hardware"?

 

there are 2 "shot" demos within the demos dir. just type "playdemo shot_demo001" or "playdemo shot_demo002" and you get the same fov of the same scene each time ...

 

I have made the difference test my own. Made a screenshot with the mod and one without, layered them with a difference filter and applied a gamma filter +5 to the result .. here is, what came around ....

http://korgano.dyndns.org:4242/~daerbaer/difference + Gamma 5.jpg

 

....unbelievers....

 

You must also define shader so...

"mtr_light_shader" ""

change it to this:

"mtr_light_shader" "textures/particles/plasmalight"

and you are done.

-------------

So does Humus tweak lower performance in some cases compared to original? Because average frame rate can be higher if certain places higher frame rates compensates enough lower frame rates in some other places.

 

Why should they not be honest? My "history":

Cat 4.7: 45.0 fps
Cat 4.9b: 48.1 fps
Cat 4.9b & Humusfix: 58.2fps

(1280x1024, Ultra Quality, 2xAA, 8xAF. AsusX800Pro hardmodded to XT, Athlon64 3000+)

 

But you have to be carefull about what you say.

Although I think what Humus has posted is particularly interesting, I think it would have been better if he

• had not said "our hardware"
• emphasised that it was an experimental hack that potentially could be incorporated into an official D3 patch.

BTW I was also thinking that some "sat" modifiers would be needed in the normalisation calculations.

I also was wondering whether the LUT texture was actually implementing a Power fuction? It could well be the case that it is some special hand-tweaked function that the artists in id found more pleasing <shrug>.

 

I don't know the performance behaviour of ATI's HW but if you've reduced the specular power to 4, perhaps the compiler replaces the "pow" with a pair of muls? <shrug>

 

So let me get this right

On r3xx hardware people are getting an average of 1.5-5fps increase with this. (Me sulks I wanted faster for my 9800 so I could have a lan game over my network when the game finally comes out here! )

on r4xx people get up to 10fps with this? (about 18-20% max)

So where is the 40% from? was that just from just one point in the game?

(Sorry if that is wrong I’ve only just got up but that is about what I see from reading the thread. )

 

Lets continue this discussion at this thread http://www.beyond3d.com/forum/viewtopic.php?t=14907

There is already rocket launcher tweak in there.

 

If I were to take a random guess it would probably be physical card limitations. There is only so much we are going to be able to squeeze out of our 9800s and the gains with each improvement will likely be exponentially less without significant reworking of driver or game code. A few little tweaks here and there just aren't going to cut it for huge gains.

In the case of the R4xx chips, they have a bit more breathing room so they stand more to gain from optimizations initially.

Right now, if you are on the lower end of the spectrum hardware requirement wise, even a small gain is a huge one. If you are on the top end, small changes can net you potential big gains.

And, for those of us in the middle to lower-top end spectrum, the small boosts we can get help, but we can't get the huge boosts we need to take it up a notch. So we kind of feel like losers all around unless you are a number fanatic

 

do you really say its the same image quality ...

... using specifically created specular texturemaps
... and just using a constant power of 16

?!

have you ever seen specular light texture maps

:?

i am sure that's faster though and i am sure now there is REALLY no difference between HIGH and ULTRA quality level in D3 anymore - but ATI / id should for sure not do it that way in their software on default

 

I presume you are thinking of homous because humus is:

 

spectacular lightmaps? I'd sure like to see those!

 

What are you using to get the fps? Timedemo1?
I had about the same improvement you did going from the 4.7b's to 4.9b's, but the Humus tweak only gave about 1.1fps extra.

I wonder whats holding it back on my PC ?
Prescott 3.06ghz, 1 gig PC3200 & Sapphire X800PRO..
Some get 10+ fps, while others only get 1-3 fps...certainly odd
BTW do you get the same increase in frame rate when using High Quality ?

 

Does this do noting for the 6800's,I think they were fast enough anyway. I just wondered. why it wouldnt?
I have a 6800gt as well btw

Oh and JC may be a bit busy his rocket blew up see here

 

Yes, I run it two times and look for the score from the second run.

Yes, in high quality also a 10fps gain. (61,5)