Check out this thread

[Keysplayr]

Some people will ab-so-f'ing-lutely do anything to discredit something their not happy with.

 

[beserker15]

alright, i've tried the mod...not sure if it's 40%...but it sure is ALOT better....no iq loss that i can see, nor artifacts. i have a softmodded 9500 which used to run the game at 1024resolution with medium setting, no AA or AF and gets a max of 60fps, most of the time it's 40-50fps, and dips to 28fps at lowest. NOW!!!...i can get 1024resolution with high setting, 2xAA/8xAF, and it stays at 60fps almost always, sometimes dipping to 34fps at lowest...

 

[Rent]

Originally posted by: keysplayr2003
Some people will ab-so-f'ing-lutely do anything to discredit something their not happy with.

Thats the internet for ya.

To those people, STFU n00bs

:laugh:

edit - 56k warning on that link

 

[Curley]

Hmm, I read the "source," and it talks about a tweak to improve performance. It does not, however, strongly imply deliberate sabatage on the part of Carmack. That implication was all yours. Therefore, sorry, you're quite flamable for what you ADDED to the source info.

Xentropy

I could have worded it better, but "deliberate sabatage" is very strong. I am impling an oversight on the attention given (or lack thereof) to ATI performance.

I implied only what Humus implied, that he can't believe how someone as smart as Carmak could miss this.

It looks as if this will be a full blown interview with JC.

I want to see what Valve Software could have done to boost NVIDIA performance in HL2.

It could have been tit for tat. Quid pro quo. A settlement.

Yes, I am pushing the conspiracy theory.

At the end of the thread, you see that they are throwing around the fact that ID or Activition has the final code etc....

I may be on fire later but right now there does seem to be something to it.

 

[ponyo]

OP post is nothing but flamebait. The link was good but everything else has no merit.

 

[KevinH]

Good info but your title is utterly ridiculous. No one is going to even give this a glance if you didn't make the title seem like typical fan boy tripe. I own an ATI card and am hoping I can get this to work but jeez you do realize you jsut managed to kill any sort of intellectual debate on this topoic here on AT when you sound like a fan boy.

 

[rbV5]

I seriously doubt there is any deliberate shenanigans to reduce performance on ATI cards. Interesting work by Humus though, will have to see how it all plays out.

 

[Juice5557]

Even if Doom 3 Was intended to make ATI Cards run slower, HL2 Is being intended to make ATI Cards run FASTER, So why is everyone complaining? Companies do this all the time

 

[KevinH]

Okay just did a a before and after. 34.5 @ 10x7 Med Qual with my 9700 Pro on two runs. With the final changes I got 36.6 fps. A negligible bump for me at least. Oh well, free frames are free. If it was 40%, I'd be doing backflips as this'll save me an upgrade .

INterestingly, I bumped it up to 10x7Q and got 35.5 on two runs. This is after the changes. Too bad I didn't do a run on HQ before the changes. REgardless, I know I can run it on HQ without taking any performance hits.

edit x 3 - Bah been testing more and the results are up and down by a few fps. I can't really tell if I made the changes correctly anymore. Can someone here give a step by step instruction on how they managed to extract the final changes and made it "stick"?

 

[BFG10K]

I'm really not sure what to think but my gut feeling is that JC would never do something like this on purpose. I'll be interested to hear what is his response is.

 

[StrangerGuy]

JC is a very respected game developer and I don't think he will stoop down to this level to cripple ATI card's performance on intent.

 

[Childs]

Originally posted by: Curley
I'll flame myself if this is false. I really hope it is. This would have long reaching implications as bad as insider trading as far as I'm concerned. To reduce sales of a certain product through manipulation is only going to hurt us, the consumer. They should have released the game to legitimate sites to do benchmarking rather than have ID "control" the software and show that NVIDIA blows ATI away.

It may be the other way around with ATI and Valve Software with the Half Life 2 benchmarks. Who knows? But it should be investigated and stopped now.

What do we have to do as consumers? Buy one of each and make our own determination or hack the game codes to get the performance $500.00 video cards are supposed to give us.

Some Quotes from B3d:

Reverend: Basically, JC wants to have the ARB2 path (his preferred path) to match the other paths (NV10/NV20/R200) for certain stuff.

Humus: I'll add that this makes a lot of sense too, since it could be an artist hell to make it look good if the specular behaves different on different platforms. So while POW is closer to the original Blinn model, this approximation looks good enough and is consistent with all other paths. I would imagine that in a future engine where he don't have to care about NV10 level hardware he would use the real POW.

Your choice of topic titles has much to be desired. Its nothing but flaimbait. Specific code paths for different cards was something that Carmack has said he wanted to avoid. Optimizations of this nature can probably be done for ATI and Nvidia, but then you'll end up with different code paths doubling QA and support. Seems like a mess best avoided, as the common gains seem to be around 5-15% for users, which could be gained just as easily by lowering resolution/quality settings. It also breaks internet play with pure servers.

I understand the alure for increasing framerates, but to question JC's integrity seems out of line.

 

[Avalon]

Welp, I gave what Humus said a try, just for giggles. I made the necessary shader changes in the code and applied them, but received no boost. 1024x768 at medium detail running timedemo1 three times for both before I applied the extra lines of changed code and after. I got 42.0fps every time. Oh well, was worth the fun in trying.

 

[KevinH]

Avalon are you overclocking your "9700 pro"? Man, I don't get anything close to what you are...

I have my XP1800+ @ 2166 as well ...

 

[Looney]

This is JC response to Humus:

I got a response from Carmack:

Quote:
Our specular function isn't a POW function, it matches the bias / scale calculations done on the NV10/NV20/R200 paths.


This of course explains why it doesn't look exactly the same. This is also a good thing. Now we don't even need to use a POW function, but we'll be fine with a MAD_SAT, which make it even faster. Got 21% boost in the timedemo with this. Now I still don't know exactly what exponent he's approximating, or if he's using different lookup tables for different materials, but I've tried this and it seems to look the same, but our fellow artifact hunters may want to verify that.

Code:
MAD_SAT R1, specular, 4.0, -3.0;


If that causes artifacts for someone, maybe approximating a lower exponent will do it:

Code:
MAD_SAT R1, specular, 3.0, -2.0;

 

[Looney]

Originally posted by: Acanthus
Theres IQ loss, its discussed all over that thread.

Um i'm not following that thread, but the one at Rage3d, Humus has come up with a solution for the secular artifacts (and which JC himself said he couldn't find any artifacts)... and i would bet the solution is on Beyond3d as well.

Here's the Rage3d thread: http://www.rage3d.com/board/showthread.php?t=33774493&page=1&pp=30

The updated is on page 3 or 4 i believe.

 

[Looney]

Yes it worked for me... the original gave me artifacts too, but it wasn't as bad as most people... you really did have to look to see them most of the time. But with the new solution, absolutely no artifacts whatsoever (just came back from a 3 hour session).

This is the updated tweak: http://esprit.campus.luth.se/~humus/temp/doom3PerformanceTweak.rar

 

[nemesismk2]

Originally posted by: StrangerGuy
JC is a very respected game developer and I don't think he will stoop down to this level to cripple ATI card's performance on intent.

 

[KevinH]

Originally posted by: Hardcore
Yes it worked for me... the original gave me artifacts too, but it wasn't as bad as most people... you really did have to look to see them most of the time. But with the new solution, absolutely no artifacts whatsoever (just came back from a 3 hour session).

This is the updated tweak: http://esprit.campus.luth.se/~humus/temp/doom3PerformanceTweak.rar

Without using AF I seem to be getting no performance boost. If anything, it seems like I"m taking NO hits for accesing 8x or 16x AF (~34-36.5 fps with the entire gamut of AF's).

 

[CaiNaM]

it's simply an inflammatory title to a post which spreads FUD by taking things out of context...

jc simply didn't want to code seperate paths for different hardware, rather std arb2 as he stated he was going to do some time ago. does ati gain performance? sure.. a couple fps in most cases; apparently not really enough for jc to justify increased coding, q&a, etc. to code different paths. it's not at all that big of a deal - and certainly not enough to accuse jc/id of bias.

 

[imported_tss4]

Originally posted by: KevinH

Originally posted by: Hardcore
Yes it worked for me... the original gave me artifacts too, but it wasn't as bad as most people... you really did have to look to see them most of the time. But with the new solution, absolutely no artifacts whatsoever (just came back from a 3 hour session).

This is the updated tweak: http://esprit.campus.luth.se/~humus/temp/doom3PerformanceTweak.rar

Without using AF I seem to be getting no performance boost. If anything, it seems like I"m taking NO hits for accesing 8x or 16x AF (~34-36.5 fps with the entire gamut of AF's).

If you have it on high, The 8x AF will override any control panel settings you use. So you won't see any difference. Not sure if the No AF of medium and lower over rides the control panel settings.

 

[ShawnD1]

Originally posted by: Hardcore
This is the updated tweak: http://esprit.campus.luth.se/~humus/temp/doom3PerformanceTweak.rar

That tweak isn't any better than the original tweak. To test the new tweak, I undid the original tweak (back to original settings).

before
first: 29.2
second: 34.4
third: 34.6

after
first: 35.4
second: 39.9
third: 40.0

40 fps is what I got with the original tweak.

 

[DarkKnight]

I lost .2 fps by using humus's tweak, didnt help me at all.

 

[ShawnD1]

Has anybody tried the tweak with a Nvidia card?

 

[Avalon]

Originally posted by: KevinH
Avalon are you overclocking your "9700 pro"? Man, I don't get anything close to what you are...

I have my XP1800+ @ 2166 as well ...

Running at stock 9700 pro speeds with my 9500, at 324/310. I tried the new tweak that hardcore posted. I went from 42FPS original score, to 42FPS with the original tweak, to 41.3FPS with the new one that he posted. This isn't doing squat for me :|

*Edit*
41.5FPS when I comment out MUL result.color, color, fragment.color; and uncomment MUL result.color.xyz, color, fragment.color; at the bottom for vertex coloring. Again, I'm getting nothing

WTF?!
I ran the timedemo again, except with settings of 12x10 on high detail with 2xAA/8xAF, just for giggles, and I scored 44.3FPS. I thought something was wrong here, so I checked my monitor while running another loop of the demo, and it said I was in 800x600. Does the demo naturally run like this, or did the game default me to it when I tried testing at a higher res?

Yes, it appears the game defaulted me down to 800x600. Hmm, that was nice and playable with the eye candy ;D

10x7 high detail 2xAA/8xAF = 31.9fps, wewt!