XGI App Detection [Archive]

Pages : [1] 2 3

StealthHawk

12-18-03, 11:32 PM

http://www.hardware.fr/html/news/

Anyone want to give the translation a go? Looks like XGI is monkeying with UT2003 quality.

They're also detecting the WHQL certification program :p (does anyone else find that hilarious?)

Lecram25

12-19-03, 01:08 AM

StealthHawk

12-19-03, 02:29 AM

Originally posted by Lecram25
*sigh*

This is so old...

tertsi already pointed this all out quite some time ago...

*cough (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=21765&pagenumber=1)* and *cough (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=21765&perpage=20&pagenumber=2)

And anyways, XGI already replied about these so called "optimizations"...

*and the third cough (http://www.xbitlabs.com/news/video/display/20031218064149.html)*

Sorry, but no. Have you looked at the news entry at the listed website where there are two UT2003 screenshots? One is with the exe named to something else and one is with the normal name exe. Proving that the drivers are detecting at least UT2003 and changing their behavior when that file is run.

Smokey

12-19-03, 03:08 AM

Well after changing the 3dmark03 exe to 3dmurk03, the mother nature scene went from 20.6fps to 8.8fps, I would say thats a pretty big drop just there!!

Johnmcl7

12-19-03, 08:11 AM

Very rough translation:

After the new ones concerning the quirks in the drives XGI, we did some essays while modifying the file of the drives XGI so that the applications more is not recognized (modification of the character chain 3 d m has r k 0 3. e x's e in 3 d m u r k 0 3. e x's e for example) since the simple file renommage notAllowed not to bypass the detection. The results are rather impressive, since for example one passes of 20.6 fps to 8.8 fps in the scene Mother Nature of 3DMark03 (1024*768, global score follows it same tendency), or again 88.2 to... 18.7 fps in the bench BotMatch integrated to Unreal Tournament 2003 (1600*1200). One will notice that the decrease is nevertheless least in our own démo, more realistic, since score does not pass « than » of 28.6 to 14.6 fps...

D’où comes this speed decrease? Specific D’optimisations to these applications well heard. The latter are can be multiple, but the verification for example of pop video l’intégration glides for the démos (as NVIDIA in 3DMar03 to an era) n’est not vérifiable without helps developer. Nevertheless, we noticed that when to drive it detected one of the listed applications, it deactivated the filtrage trilinéaire for do only of the simple bilinear one, c'est-à-dire without no transition between the different levels of mip map. Well heard that reduces the work load by two (interpolation from 4 texels instead of 8). Some does being given the importance of the decrease one can even to wonder if the second GPU well activated "by default"...

Here to illustrate our matter a screenshot under Unreal Tournament 2003 with the drives d’origines, and another screenshot once these detections are deactivated:

We here used the order firstcoloredmip for that the different levels of details is colored in order well to make obvious the difference at the level of the filtrage.

Worse, under Halo with the detection d’origine, the graphic quality is really deplorable, has such not at all qu’on would believe itself in 640*480 when one is in 1024*768, so well that l’on thinks qu’il not s’agit uniquely d’un problem of filtrage trilinéaire. This problem is determined when l’on deactivates the optimisations, but suddenly one attends a slide show. It is to note that the function screenshot does not work correctly when Halo is detected by to drive it, this that prevents us from do you to see the quality difference graphic.

Here that explains therefore largely why in the benchs running the Volari s’en goes out comparatively well, and why in games less running for the but benchs not necessarily less played c’est the catastrophe. In our test protocol habitual using in major party of the non detected games, the Volari Duet V8 Extremist arrived in fact to 55% of performances dune GeForce4 Ti 4600, with in the better ones of the cases 77% of the performances of this last one under UT2003 and the specific optimisations that go with (40% without) and 15% of better under Quake 3, since in OpenGL by default and oppositein Direct3D XGI deactivates office the filtrage trilinéaire and lowers equally by default the detail level of the textures as you can note it while comparing the ati filtrages and of XGI:

To the final one, one can ask oneself which is l’intérêt for XGI d’intégrer this kind of things in these drives. In fact, if XGI thinks that its fleas are not rather quick to carry out a filtrage trilinéaire, as much to deactivate it totally drives rather than to deactivate it that in certain applications that generally are used to measure the performances of the graphic cards. It does not remain anymore qu’à to hope for XGI qu’il will learn from its errors and than of futures drive will come to correct the shooting and will improve strongly the performances as the quality in order to have a worthy result dune card to 400 €...

bkswaney

12-19-03, 10:31 PM

At this rate I would not be supprised to see every company start doing this.
But as long as IQ is good I could care less.

Kyro chips do the same thing. Each game is fine tuned to run it's best.

It can be a good thing. Just as long as IQ does not suffer.

That's my 2 cents anyway. ;)

NickSpolec

12-20-03, 12:47 AM

Kyro chips do the same thing. Each game is fine tuned to run it's best.

Umm.. No, they don't.

Each game might have one or two compatibility settings in the registry, but that's it. They are not "optimizations" for the sake of fooling gamers with lesser quality/high frames.

bkswaney

12-20-03, 01:06 AM

Originally posted by NickSpolec
Umm.. No, they don't.

Each game might have one or two compatibility settings in the registry, but that's it. They are not "optimizations" for the sake of fooling gamers with lesser quality/high frames.

Does anyone know this for sure?
I do not recall ever seeing it investigated.

Malfunction

12-20-03, 01:19 AM

Originally posted by StealthHawk
http://www.hardware.fr/html/news/

Anyone want to give the translation a go? Looks like XGI is monkeying with UT2003 quality.

They're also detecting the WHQL certification program :p (does anyone else find that hilarious?)

Originally posted by tertsi
Here is possibly Directx/D3D app-detections (found in xrvkp.sys) ....

OVERDRAW.EXE
3DMARK2000.EXE
3DMARK2001
UNREALTOURNAMENT.EXE
TRUCKRACE.EXE
GIANTS.EXE
COHOKUM.EXE
ODYSSEE.EXE
IF18.EXE
STEAMBOAT.EXE
DEVIL.EXE
ROADWAR.EXE
STUNTGP_D3D.EXE
GAME.EXE
RF.EXE
GROUCH.EXE
TRIBES2.EXE
AQUANOX.EXE
DUNGEONSIEGE.EXE
RALLYTROPHY.EXE
MRALLYE.EXE
SPIDERMAN.EXE
NASCARRACING 4.EXE
F1.EXE
SERIOUSSAM.EXE
4x4.EXE
MAXPAYNE.EXE
UW.EXE
NR2002.EXE
CHAMELEONDX.EXE
4X42.EXE
BEND3DIM.EXE
UT2003.EXE
HH2003.EXE
MORROWIND.EXE
SIM.EXE
ARMYOPS.EXE
LITHTECH.EXE
SIM.EXE
NWO.EXE
EVOLVA.EXE
SR2.EXE
F1_2002.EXE
BF1942.EXE
AQUA.EXE
FIFA2002.EXE
NFSHP2.EXE
HITMAN2.EXE
AQUAMARK.EXE
IRONSTORM.EXE
SKATE3.EXE
VIETCONG.EXE
3DMARK03.EXE
GAME.DAT
DFBHD.EXE
SPLINTERCELL.EXE
DFBHDD.EXE
INDY.EXE
FELLOWSHIP.EXE
WAR3.EXE
MFCTEX.EXE
MOIRE.SCR
GMMARK2.EXE
UT2003.EXE
MADDEN04.EXE
HALO.EXE
3DNASYS.EXE
TRAOD.EXE
TRAOD_P3.EXE
TRAOD_P4.EXE
SEP.EXE
CLIENT.EXE
DFBHD.EXE
MGS2
TOEE.EXE
MAJESTICCHESS.EXE
AOWSM.EXE
F1_2001.exe
SPEEDDEMO.EXE
X2-DEMO.EXE
XIII.EXE
FFXIWINBENCH.EXE
SIMCITY4.EXE
SM.EXE
FREELANCER.EXE
NR2003DEMO.EXE
FS9.EXE

Wasn't this sorta already known about awhile ago after looking into their 30+ meg driver?

http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=21765&perpage=20&pagenumber=2

Peace,

:confused:

The Baron

12-20-03, 01:32 AM

Wavey Dave confirmed it by completely discrediting the site that's saying that the app detection is only for compatibility...

http://www.beyond3d.com/forum/viewtopic.php?t=9481&postdays=0&postorder=asc&start=40

StealthHawk

12-20-03, 04:35 AM

Originally posted by Malfunction
Wasn't this sorta already known about awhile ago after looking into their 30+ meg driver?

http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=21765&perpage=20&pagenumber=2

Peace,

:confused:

I really don't know why I should have to say this twice, but here goes.

The point of this post was not to say "look, XGI has application names in their drivers, maybe they are doing something." The point entirely was to say "look, ut2003.exe is one of the application names in the drivers, and lo and behold the drivers are doing things differently when ut2003.exe is run and when the drivers are hex edited and the ut2003.exe reference is changed to something else so it is not detected."

Concisely, we have a claim here that something is going on, IQ is changing in UT2003, performance is dropping big time in GT4 of 3dmark03. This is not speculation anymore.

Read Johnmcl7's post for a translation of the newspost...

Skuzzy

12-20-03, 04:50 AM

Originally posted by StealthHawk
I really don't know why I should have to say this twice, but here goes.

NVidia has done their marketing well, would be my guess.

sxotty

12-20-03, 12:10 PM

SH didn't you know XGI isn't nvidia, so it is fine for them to cheat... sheesh how complicated is that.

Lecram25

12-20-03, 01:02 PM

It seems that only Hardware.fr are the only ones who seem to find these "cheats"... xglider from volarigamers and the guys from hardwareteaker both changed file names and didn't see any performance issues...

The Baron

12-20-03, 01:41 PM

Originally posted by Lecram25
It seems that only Hardware.fr are the only ones who seem to find these "cheats"... xglider from volarigamers and the guys from hardwareteaker both changed file names and didn't see any performance issues...
check the link I posted about volarigamers. In other words, it's got more than a few remarkable connections to XGI PR.

StealthHawk

12-20-03, 02:44 PM

Originally posted by Lecram25
It seems that only Hardware.fr are the only ones who seem to find these "cheats"... xglider from volarigamers and the guys from hardwareteaker both changed file names and didn't see any performance issues...

Did they just change the name of the files?

Or did they change the file name reference in the drivers, as that is what is causing the performance drop according to Hardware.fr.

digitalwanderer

12-20-03, 03:09 PM

Originally posted by Lecram25
It seems that only Hardware.fr are the only ones who seem to find these "cheats"... xglider from volarigamers and the guys from hardwareteaker both changed file names and didn't see any performance issues...
DaveBaumann at that link Baron posted:
However, considering there are no Volari products I have difficulty in accepting that Volarigamers is a site set up by independant enthusiasts. The fact that both XGITECH.COM and VOLARIGAMERS.COM were both registered through the same company (GoDaddy.com) and Volarigamers contact details are "Domains by Proxy, Inc." doesn't do anything to asuage my suspicions. It may well be, but then I've not really seen anything to suggest that isn't the case.
Basically the two saying it's no biggy are owned by XGI, a bit suspect and it makes 'em a bit less credible.

TheOneKEA

12-20-03, 03:15 PM

Ugh ugh ugh ugh ugh.

I'm slowly losing my initial desire to wait for an nV or [H] review - I'm beginning to think that XGI <- ****e, Volari <- ****e and that the world of high-end video is still bipolar.

Lecram25

12-20-03, 03:44 PM

I must stand corrected...the way hardware.fr checked for this "cheat" was via hex-editing the .exe recognition from the drivers which in turn lead to the big performance drop...maybe XGI did this so that the drivers that would be/were released with the card would make the card competetive to the big dogs out there, whilst they work on a much better driver set...the ones that are meant to be released sometime this week :S

digitalwanderer

12-20-03, 04:25 PM

Originally posted by Lecram25
I must stand corrected...the way hardware.fr checked for this "cheat" was via hex-editing the .exe recognition from the drivers which in turn lead to the big performance drop...maybe XGI did this so that the drivers that would be/were released with the card would make the card competetive to the big dogs out there, whilst they work on a much better driver set...the ones that are meant to be released sometime this week :S
I honestly hope you're right as I want another player to enter the graphics arena and not another lamer, but I am NOT encouraged that they'd do this even for initial launch impressions with beta drivers. :(

Man, they should have just SO known better with all the flack nVidia is getting...I find it ridiculous. The only reason I can see them doing it is if they REALLY had to do it since their hardware under-performs incredibly, but that ain't good news either. :(

The Baron

12-20-03, 04:30 PM

1. Launch reviews are the only reviews that really matter. Case in point (or so I'm told repeatedly whenever I bring it up): R200. With the exception of the Volari now, it was THE WORST LAUNCH EVER.

2. Since only launch reviews really matter to people, performance on launch samples (or samples within 2 weeks of launch) must be good through any means necessary.

3. Volari is a definite underperformer at this stage. Maybe it'll be improved through drivers, but I think it's just another case of two chips without any efficiency whatsoever.

4. XGI adds cheats to create the illusion of good performance in launch samples and will probably remove the bulk of them if and when drivers increase performance significantly.

Lecram25

12-20-03, 05:10 PM

I'm waiting for the new drivers to be released, and that will be XGI's final ulitmatum in my eyes...

ChrisW

12-20-03, 08:21 PM

Too little, too late. Perhaps they can fix their problems with future drivers. The problem is it is being released at the end of the product cycle and the R4xx and NV4x will be released soon. Who is even going to care about this card when those cards are released? So what if they can finally reach R3xx and NV3x speeds after the next generation of chips are released.

Lecram25

12-20-03, 08:42 PM

AFAIK, the XG45 (new chipset meant to compete with the nv40 and r420) will be out around the same time as the new upcoming chips...the Volaris that we're "seeing" now were sort of a marketing ploy to create awarness and a way for XGI to make a name for themselves, and some money too...looks like that kinda failed, in both aspects...

The Baron

12-20-03, 09:13 PM

Originally posted by Lecram25
AFAIK, the XG45 (new chipset meant to compete with the nv40 and r420) will be out around the same time as the new upcoming chips...the Volaris that we're "seeing" now were sort of a marketing ploy to create awarness and a way for XGI to make a name for themselves, and some money too...looks like that kinda failed, in both aspects...
We're not seeing driver problems alone. We're seeing once again that multi-chip simply is not feasible if you want any sort of reliability, profit margin, or yield.