My thread @ Futuremark(Re =Waite for Unwinder)

Regarding the game benchmark application detection for both Nvidia and ATI, how do we know that these are not legitimate optimizations? While optimizing for 3dmark clearly defeats the purpose on synthetic benchmark, optimizing game performance is more then welcome as far as I am concerned, as long as IQ is not adversely effected.

 

Why can't they make general optimizations instead of application specific ones then?

 

Sometimes the game authors do stupid things and aren't willing to patch it after the fact.

 

My apologies, the previous screenshots on NV boards were really grabbed with AF on. Updated AntiDetect.rar will be uploaded few minutes later.

 

Geeforcer,

if you look at the other threads here you can see that these optimizations only seem to apply to popular benchmarking programs. Also factor in that when you run a different timedemo (such as FS did) you come out with much different results than using the standard timedemos. If there optimizations are legit for games then they should apply all the time that the game is running and not just for the standard time demos.

 

It's no secret that IHVs have been optimizing drivers for the more popular games since the early days. Again, as long as long as IQ is not effected, I don't particularly care what they do; I'll take the extra FPS. Now, if these optimizations only boost scored in a particular demo and do not increase performance anywhere else in the game, then it's cheating as the IHV is simply trying to misrepresent its card's overall performance in the game

By FS you mean firing squad? With all the "Cheating" threads around I am heaving trouble finding the one you are referring to, could you please link me? Thanks.

 

well, i for one remember, when ati was caught on quack deal, they publicaly stated they will be optimizing their drivers for quake3 engine, so performance could be spotted in all q3 driven games (jedi knight, elite force, etc). as far as i am concerned, this is legit and welcome.
usual caveat: as long as IQ is not compromised.

so, here is my question. can anticheat tool also find those optimisations (yes, i know it focuses on d3d for now, and quake is opengl) and more importantly, are those still valid optimizations or cheats? WE can also expect more unreal engine driven games, so, if optimizations for ut2k3 show in other games, too, isn't that a valid optimization for popular engine?

it is just like optimizing for autocad or 3DS max, isn't it?

 

Oh I agree. And I am sure there are ton of perfectly legit optimizations that nV has made. I am not one to stand in the way of progress. More power to them as long as they are legit. Thats seems to be the issue. Look at the GamePC review where they used less common games:

http://www.beyond3d.com/forum/viewtopic.php?t=6499

Sure,
http://www.beyond3d.com/forum/viewtopic.php?t=6459

BTW I have trouble keeping up on them as well... But one you look at all the results and how thier optimization seem to be targeted at benchmarks (code creatures, 3dmark, Aquamark, SS, UT2k3 and CS). And thats not really a good thing.

 

No, depending how you optimise, then it may not apply to other games based on the same engine. EG, UT2k3 specific pixel shader replacement wouldn't necessarily be applicable to other UT-engined games (unless, of course, they are using the exact same pixel shader).

Of course, taking advantage of camera rails and static clipping planes (if this is going on elsewhere than 3DM03...) will never even apply to playing the game itself, much less other games using the same engine.

 

Unwinder

So this is just further investigation of of the 3dmark incident ? Well it's nice to know that nvidia just cheated again but I don't see what your getting at with ATI, I would suggest that you try your test against maybe 4-5 driver versions of each card vendor, comparing IQ all the way and releasing those pics and techniques used after everything is done with. This would definately ground your work in integrity, allowing others to reproduce the results so their is no doubt. As of right now this seems to do nothing more but attempt to say ATI(damage their character) did more what than they admitted to without concrete proof. Also throw a few game benches in their, please

 

On russian:
http://www.ixbt.com/video2/antidetect.shtml
Antidetector scripts:
http://www.ixbt.com/video2/images/antidetect/antidetectorscripts.zip
Results in 3dmark2001
Game4 normal/with AntiDetect
GF4600 75/43 (changes vertex shader)
R9500Pro 66/47 (changes texture format&compressing)
R8500LE 42/31

So... waitng for reviews with AntiDetector scripts....

 

Hmh, I thought Digit-Life was supposed to post an article about this? Now it seems that the writer of this script has posted it himself. So, didn't Digit-Life think this thingie is big enough to go public or are they still writing their report?

 

AFAIK testing is almost done, now writing&preparing
Edit: It seems the order of publications will be:
1. 5900&3dmark2003 without anti-detect & 44.65
2. 5900 vs 5900U vs R9800Pro
3. Anti-detect - comparing how much cards lose in different games&demos

 

thanks for FYi...

hopefully the english translation will show exactly what is going on and exactly WHAT unwinder did to display this information...

will be nice to eventually see higher end cards from both ihv's tested and info released as to whats going on

 

This whole issue seems to be very complicated (atleast for me), so thanks everybody for your intersting comments on the matter.
Also thanks Unwinder for your effort.

I'm not sure if Unwinder's detection mechanisms catch the AF improvement witnessed by renaming the 3DMark03.exe (the 3DMurk issue), so you might do that additionally when benchmarking with AF on.
If I'm wrong feel free to correct me.

Keep up the good work!

 

I actually completely agree with this. A benchmark is one thing--and frankly, even in a benchmark as long as no reduction of the default workload is achieved, optimizations don't bother me in the least. (Reducing the workload to simulate optimization, however, would be cheating.)

But in a game--I don't think it matters... Do what nVidia did in 3DMk with clip planes and so forth and you'll produce a set of drivers so buggy the game will be unplayable--so that kind of thing isn't a worry in real 3D games. Games are meant to be played, not used as benchmarks. If an IHV does something like sacrifice image quality to inflate performance in a game--that's where the role of the thorough hardware reviewer comes in and he can spot that when it happens and make people aware of it.

 

English version of this article

Will be published today, later, at http://www.digit-life.com

 

Re: English version of this article

WOO-HOO! Thanks for the heads-up. 8)

 

Tomorrow, then? I hope this isn't one of those nVidia "Are You Ready--Psyche!!1(TM)" hoaxes.