ati and ut 2003 "cheat"

O

olivier

Newcomer
saw this at hardware.fr

use your favorite translater

En fait, cela est lié à la gestion du LOD (Level Of Detail – Niveau de détail) des textures chez ATI et à l'utilisation de ce paramètre dans Unreal Tournament. En effet, afin d’améliorer les performances des cartes graphiques sous UT2003, Epic utilise dans Unreal Tournament un LOD positif pour ces textures détaillées. Il est de 0.8 par défaut (DetailTexMipBias dans UT2003.ini), ce qui signifie que les textures passent plus rapidement (à une distance plus proche) au niveau de détail inférieur (mipmap) que cela se fait normalement (LOD à 0).

Seul problème, à partir d’un LOD de 0.5, les cartes ATI n’affichent plus du tout le premier niveau de détail de textures, le plus précis, lorsqu’on utilise l’anisotropic filtering alors que chez NVIDIA c’est encore le cas du fait d’un algorithme de calcul du LOD différent. Du coup, d’un point de vue qualitatif NVIDIA offre un meilleur rendu sur les textures très proches du point de vue avec ce réglage à 0.8.

S’agit t’il d’un "cheat" ? Pour en avoir le cœur net c’est assez simple, il suffit de tester ce que sont les performances avec un LOD configuré à 0.4 dans UT2003, ce réglage donnant les mêmes résultats chez ATI et NVIDIA. Sur une Radeon 9800XT la baisse constatée est de ... 0.9%, et elle est de 2.3% sur un GeForce FX 5950. Passer carrément à un LOD 0 sur ATI ne baisse d'ailleurs pas plus les performances. Bref, si il s’agissait d’une optimisation de la part d’ATI, elle serait bien mauvaise. Reste une question ... qu’est ce qui a poussé Epic à monter au créneau à côté de NVIDIA alors qu’un simple paramètre dans la configuration de leur jeu permet de résoudre le problème ?
Click to expand...

they said the "cheat" give ati a boost of 0.9% and we can resolve the "cheat" with a parameter in the config file of ut 2003.
 
How can ATI be cheating if when you disable DetailTexMipBias & DefaultTexMipBias in the Ut2003.ini the image quality goes back to high quality thats better than your competitors, yet with the commands active they force the lowest quality d3d level possible - also why is the same command in opengl set to high quality?

www.jlmay.f2s.com/UT2003 Anubis.jpg

If I write a game and add a command that deliberately slows down all AMD CPU based PC's, who's in the wrong AMD or me for adding such a command?

Also has anybody checked to see if the commands effect NV FX range?
 
I find it amazing that anybody can call .9% a cheat with a straight face. That's so far in the noise its not even funny.
 
I don't know why they even bother during a fps comparison for this issue. It's a texture problem and it's noticable. Whether or not it increases the framerate or not is moot. The fact is, it's not drawing all the textures in high-res, which is a problem...
 
RussSchultz said:
I find it amazing that anybody can call .9% a cheat with a straight face. That's so far in the noise its not even funny.
Click to expand...

Oh my, that's a .3fps change (from a 100fps baseline)! WHATEVER SHALL WE DO?! ATi gained .3fps! Wow, that's gonna change people's opinion of the benchmark results...
 
Kalbaz said:
I don't know why they even bother during a fps comparison for this issue. It's a texture problem and it's noticable. Whether or not it increases the framerate or not is moot. The fact is, it's not drawing all the textures in high-res, which is a problem...
Click to expand...
To proove that it's not a "questionnable optimisation" (taken from tom's hardware) since there is no performance hit.
 
The question is : Is the issue a result of how Epic tuned the detailtexture lod or is the issue a result of how ATI is calculating the lod ? It is hard to answer this question.

I think the problem comes from both points. I also think that Epic can correct this easily instead of insinuating that ATI is cheating and forgetting to render the detail textures.
 
I wouldn't say this is a cheat, but I think it's a bug. This issue has been known for a looong time now (almost as long as UT2k3 has been out now), I'm really surprised that it's a) again making an impact and b) still not fixed yet.
 
Xmas said:
I wouldn't say this is a cheat, but I think it's a bug. This issue has been known for a looong time now (almost as long as UT2k3 has been out now), I'm really surprised that it's a) again making an impact and b) still not fixed yet.
Click to expand...

I fully agree. The funny part of this issue is that NVIDIA could have chosen another view for the quality comparison. It seems like some level designers of ut2003 have used the detail texture as the main texture on some objects... On these objects the texture quality is really bad with ATI boards... Maybe NVIDIA has not noticed this yet ;)
 
I wouldn't be suprised if the Nvidia cards just ignored the mipmapbias setting whereas a setting of +8 puts R3x0 cards into performance driver setting.
I don't really see "how" ATI could fix this as in the UT2003 opengl section the mipmapbias is set to 0 - which is quality mode.

To get a better understanding of why mipmapbias is set to +8 tests need to be done with an FX card to see if the image quality changes when the mipmapbias is set to lower levels (it maybe that after +2 there's no change for instance) and also test the fps performance at these various settings.
Tests should also be run with the ingame commands disabled and compare images for the FX and R3x0 with drivers to max quality and also compare the performance hit difference.
My hunch is that the D3D setting of +8 has been done deliberately to hurt non "The way it's meant to be played" cards.
 
I can't visit the site, because it makes my pc crash, but is Guru3d using Hanners test to say ATI is cheating?

this is after reading this piece of text on another forum
Next to that Elite Bastards today decided to check out the implications themselves and came to the same conclusion. Nice Job Hanners

"When looking at these images, concentrate on the foot and leg of the statue - On closer inspection you can see the difference in texture quality between the two cards - The nVidia board does indeed look like it is using the detailed texture (as it should), while the 9800 Pro looks like it is using a less detailed, lower quality version. Although the positions the two screenshots have been taken from differ a bit, I can assure you from my own testing that the symptom captured in these images is the same regardless of the position you view the statue from."
Click to expand...
 
hjs said:
I can't visit the site, because it makes my pc crash, but is Guru3d using Hanners test to say ATI is cheating?

this is after reading this piece of text on another forum
Next to that Elite Bastards today decided to check out the implications themselves and came to the same conclusion. Nice Job Hanners

"When looking at these images, concentrate on the foot and leg of the statue - On closer inspection you can see the difference in texture quality between the two cards - The nVidia board does indeed look like it is using the detailed texture (as it should), while the 9800 Pro looks like it is using a less detailed, lower quality version. Although the positions the two screenshots have been taken from differ a bit, I can assure you from my own testing that the symptom captured in these images is the same regardless of the position you view the statue from."
Click to expand...
Click to expand...
'Xcuse me, I gotta go educate a website about misinterpretting someone elses work.... :mad:
 
You must log in or register to reply here.

Similar threads

O
Sébastien Loeb visits Polyphony Digital (GT4 as the real driving simulator)
Replies
13
Views
3K
Nesh
N
P
Some Strange PS3 sale numbers
2
Replies
29
Views
5K
NERO
N
H
Welp It's Official - SCEF president: PS3 launch details at E3
2 3 4
Replies
72
Views
10K
Titanio
T
H
niNtendo Paris EURO conference. : DS 11march 149 euro
Replies
4
Views
2K
Zeross
Z
W
Battlefield: Modern Combat announced for Playstation 2
Replies
8
Views
2K
wco81
W
Back
Top