enjoy about 40% higher performance in Doom3 Using ATI cards

Gorgon

Newbie
Joined
Aug 20, 2003
Messages
6,684
Reaction score
0
http://www.beyond3d.com/forum/viewtopic.php?t=14874

Conclusion and discussion:
I don't want to complain about Carmack's work, I still consider him to be the industry leader in graphics engines. Though when I read the shader it striked me how many texture accesses it did compared to the relatively short shader, even for stuff that could just as well be done with math for a small cost in instructions. Using a dependent texture lookup for POW evaluation makes a lot of sense for R200 level hardware due to instruction set limits, but for R300 and up it's much better to just spend the three cycles it takes to evaluate POW with math instead of risking texture cache trashing with a dependent texture read, which may be much more costly, especially since the access pattern in this case will be far from linear. Also, using math improves the quality too, even though it may not be very noticable in this game.

:cheers:
 
DiSTuRbEd said:
FU NVIDIA, all I gotta say! :) :E

Jumping the gun a little bit - this 'tweak' doesn't seem to make much difference at all. Some people have reported a 1 fps increase wooohoo \o/

NVIDEA must be quaking in their boots ;)
 
People in that thread at Beyond3D have reported a loss in image quality with this "fix", so it's more like a performance/IQ trade-off than a pure speed boost.
 
Wow, I just read the entire thread, and oh my god! That's a VERY nice improvement!
 
Warbie said:
Jumping the gun a little bit - this 'tweak' doesn't seem to make much difference at all. Some people have reported a 1 fps increase wooohoo \o/

I already get 60fps almost constant, so its not like I need to improve anything, more like the other card owners.
 
40%

I think that's a bit optimistic. Anyway, has anyone tried this "fix" using nVidia hardware??

It might actually provide a boost on nVidia cards as well.
 
People who tried says no boost at all on Nvidia Hardware.
 
It makes no difference on my 9800 pro. The biggest improvement i've seen it make (aside from what the guy who came up with the idea claims) was 3 fps on a X800 pro.

So, in some cases it seems to make a very slight performance increase, on most, none at all. A million miles from a 40% increase.
 
Zerox said:
People who tried says no boost at all on Nvidia Hardware.

Everything in that file you posted looks like it's got something to do with image quality anyway, there's almost certainly a trade off. These increases (in most cases small) probably hit the image quality.
 
Um, Warbie, read the entire thread, quit a few people see 8-15 framse more per second.
 
hmm maybe ill try it my fps arent that hot on my 9700pro :flame:

but with almost all games ppl have these kinds of uber performance tweaks that never seem to work :hmph: (atleast not for me)
 
Yeah... but, you can get these increases.

x800Pro

Before:
1024x768 High NoAA/8XAF = 73.3fps
After:
1024x768 High NoAA/8XAF = 88fps

9800Pro

Before:
42.4
After:
57.5
 
mortiz said:
Everything in that file you posted looks like it's got something to do with image quality anyway, there's almost certainly a trade off. These increases (in most cases small) probably hit the image quality.

Please, at least try it before judging.
 
You have to remember doom3 is not an accurate benchmark and playing the same demo over on the same setup multiple times results in a totally different average FPS each time - hence why most people run the same demo about 10 times and then work out the average themselves.

Some people on the Beyond3D thread didn't do this, hence why some of the results seem completely random - it's very misleading.

The 'tweak' only seems to be good for an average extra 1-5 frames per second anyway, and as people say... image quality is sacrificed as the shader mod just takes rendering shortcuts.

Not worth it IMO, ID made that shader like that for a reason.
 
lePobz said:
You have to remember doom3 is not an accurate benchmark and playing the same demo over on the same setup multiple times results in a totally different average FPS each time.

I ran i several times and still got the same FPS.
 
Zerox said:
Yeah... but, you can get these increases.

x800Pro

Before:
1024x768 High NoAA/8XAF = 73.3fps
After:
1024x768 High NoAA/8XAF = 88fps

9800Pro

Before:
42.4
After:
57.5

I'd take that with a big pinch of salt at the moment. I've tried this tweak on 2 9800 pro systems and one 9700 pro system. neither have had any increase at all (one about 1 fps less :/)

Of course this may be different with the X800's.
 
I don't have a really good eye for image quality... but it has improved my performance on high quality a good bit... only ~2 fps for medium, though.
 
Wow!...I just got 2 FPS more on my 9800 pro, whoopie-de- doo. The Tweak isn't very useful, at least for me. :(
 
Humus's thing is eally quite overrated. His "40% increase" is really on visable on low end cards, where 40% is 2 or 3 fps. Alot of people who have tried this have gotten bad artifacting, and some people have had to re-install the game because of what it has done.


Yes, a lot of people are seeing improvments, but I'm not sure if it's worth it.
 
As my physics professor likes to say: Percentages are for people that never took a single math course. Of all the statistical methods, percentages mean the least. Hell I get a 100% incrase in Doom 3 when i use this tweak. But I'm playing it at 1600X1200, with everything maxed and instead of 1fps I get 2fps. Yay.

Being more serious, I get a small increase on my 9800 Pro, but nothing mind boggling.
 
Take a look:

0xAA/16xAF
1280x1024
High
9800 Pro 128MB
timedemo demo1

------------
First, before tweaks
27.5

Cat 4.9
28.9

After 40% Increase tweak
35.4

After cacheMegs = 96
38.3

After cacheMegs = 128
38.3

Use imagecache = 1 and increase minK = 20480
36.9

Refresh rate = 85Hz
37.5
---------
FINAL: 43.2 (after 3 runs, result from first: 38.2, 2nd: 43.2, 3rd: 43.2)

Holy jesus, what a nice job all the tweaks do TOGETHER :farmer:

BTW, I did these tweaks too: http://www.viperlair.com/articles/howto/software/tweakd3/

I read that you should run the demo 3 times and take the results from the 3rd, that's more true-to-life. It doesn't pause then and slow it down.
 
Yes, humus has done a good job and its working well if you know how to do it. For some reason DOOM3 even when you tell it to run a certin way till not and you need to FORCE it in the ATI panel.
 
PS
Humus work at ATI as an "Developer relations Engineer".
 
Back
Top