The Silent Hand of Driver "Optimization" (spinoff)

[Davros]

There's always been games that needed a certain driver or above to work, most famous was Alien vs Predator for about 2 years nvidia claimed it couldn't be fixed in the driver(despite it working on earlier drivers) , so a modder had a look at it found the problem and released a fix, next driver version had a profile that made the fix not needed - the take home is nvidia didn't bother trying to fix it, and just told people it couldn't be fixed
the other famous problem was the opengl extension limit bug, some games got a profile those that didnt could be made to work by renaming the exe to quake3.exe or mohaa.exe. They eventually did have an option to limit the opengl extensions reported to a game (i wonder if that option still exists)
Edit: no I guess they assume no one will play one of the affected games I doubt they've created profiles for all of them.

 

[pcchen]

It's more common in the old days because D3D spec was not as clear and sometimes old drivers had some bugs which games were designed around those bugs (or the devs probably think they are the intended behavior). Once those bugs were corrected, old games suddenly don't work correctly, especially if they are very old and no longer being maintained. Unfortunately, new games might rely on the new "correct" behavior so you can't just revert the corrections, because new games probably won't work (and, in principle, a driver should behave correctly instead of catering to some old historical incorrect behvaior anyway). It can be quite cumbersome to maintain a list of games and deliberately work in an incorrect manner. If a game is very popular it's probably worth doing it, but a popular game is also more likely to be properly maintained and wouldn't have to rely on these hacks.

 

[Davros]

My issue wasn't with the game or the bug, it was with nvidia claiming it couldn't be fixed when it was obvious they never even tried.

 

[pcchen]

As I said, the problem is when you need to make a speicifc profile for a game, what's the criteria? How popular a game should be to justify it? I mean, people are complaining that drivers are getting too big, and game specific profiles are part of the reason. Not to mention that it takes resources to do so.
As for your example, at least to me it's not hard to imagine that NVIDIA didn't want to spend resources on fixing the game (it's quite possible that it's not even their fault). Once someone did that hard work, they knew how to do it and just applied the fix.

Personally I really don't like application specific fixes, partly because I'm a software engineer myself. Poeple love to say that Windows is bloated, but one of the reason why Windows is so bloated is because they have a lot of application specific fixes, for compatibility reasons. It's already very hard to resist, I really don't think people should encourage this kind of behavior. Application vendors should fix their applications, not to rely on OS or drivers to take care of their problems.

 

[Davros]

at least to me it's not hard to imagine that NVIDIA didn't want to spend resources on fixing the game

Exactly, so why not state that, dont lie about it and claim it's unfixable. It worked with older drivers so it's not like they didn't know how to fix it

Personally I really don't like application specific fixes

Nvidia have made a rod for their own back
I've been complaining about this for years that's the reason we have graphics drivers that are bigger than the entirety of windows XP and will soon be too big to fit on a cd.

 

[pcchen]

Exactly, so why not state that, dont lie about it and claim it's unfixable. It worked with older drivers so it's not like they didn't know how to fix it


Nvidia have made a rod for their own back
I've been complaining about this for years that's the reason we have graphics drivers that are bigger than the entirety of windows XP and will soon be too big to fit on a cd.

Yet you want NVIDIA to "fix" other people's problem.
As I explained, something "worked" with older driver does not mean it's fine. I don't know exactly what's the problem with the game, but generally it's should be the game's dev to fix problems, not the other way around.
If you don't like drivers being too large, you should argue for the devs to fix the problems, not the IHV to "fix" with their drivers.

 

[Davros]

Yet you want NVIDIA to "fix" other people's problem.

No I want nvidia to not lie and claim it cant be fixed, I would have accepted "sorry it's an old title it's not worth our time to fix it"

As I explained, something "worked" with older driver does not mean it's fine.

No but it does mean it can be fixed, as does the game working fine on amd cards, intel igp's and newer nvidia drivers

 

[pharma]

Since the game worked with Nvidia's newer drivers it seems to indicate maybe it was unfixable at the time the issue was identified.
Just because a modder's hack worked means very little in relation to updating the driver/game software to provide a complete fix.

Platforms like NexusMods provides fixes to many game issues but does not mean it is a proper fix or what other issues it might create.

 

[Davros]

It was certainly fixable before the issue was identified. Other ihv's had no problem fixing the issue
unless you believe that for some period of time nvidia drivers lost the ability to clear the z-buffer

Just because a modder's hack worked means very little in relation to updating the driver/game software to provide a complete fix.

The game was updated, the modder downloaded the sourcecode found the problem fixed it and then recompiled it i'd say that qualifies as updating the software and providing a complete fix.

 

[pcchen]

It was certainly fixable before the issue was identified. Other ihv's had no problem fixing the issue

The game was updated, the modder downloaded the sourcecode found the problem fixed it and then recompiled it i'd say that qualifies as updating the software and providing a complete fix.

I'm curious, if the game can be fixed this way, why the devs didn't just apply a similar fix, but to rely on a driver fix?
Again, I don't know the exact nature of the problem, but to me it's not clear whether it's a real driver bug or just a software bug. Many times drivers have to accomodate software bugs, and I'm just saying that we shouldn't encourage this.

 

[Davros]

no bloody idea, could be that ihv's are to partly blame for allowing them to get away with it
I agree with you, I've stated ever since nvidia introduced game profiles that it's the wrong approach, they should have spent their time optimising the driver for direct3d rather than individual games.

 

[DegustatoR]

I agree with you, I've stated ever since nvidia introduced game profiles that it's the wrong approach, they should have spent their time optimising the driver for direct3d rather than individual games.

Profiles are there for game specific features and user control. The driver is generally optimized for every application, even when some of them are being launched as optimized for just one particular it is usually a generic optimization.

 

[Davros]

Speaking of profiles does anyone remember the quack3.exe scandal ?
where nvidia were detecting quake3 and lowering image quality to inflate benchmark scores caught when someone hex edited quake3.exe and replaced all instances of quake3 with quack3 and the framerates dropped (ati also did something similar) getting caught didnt stop nvidia from cheating they just used more sophisticated methods of detecting games and benchmarks like fingerprinting the camera path in 3dmark and adding static clip planes to the scene.
no ones suggesting nvidia are still doing that but then again is anyone testing for it?

 

[pharma]

Speaking of profiles does anyone remember the quack3.exe scandal ?
where nvidia were detecting quake3 and lowering image quality to inflate benchmark scores caught when someone hex edited quake3.exe and replaced all instances of quake3 with quack3 and the framerates dropped (ati also did something similar) getting caught didnt stop nvidia from cheating they just used more sophisticated methods of detecting games and benchmarks like fingerprinting the camera path in 3dmark and adding static clip planes to the scene.
no ones suggesting nvidia are still doing that but then again is anyone testing for it?

@Davros,

What is with all this conspiracy talk? If you are not satisfied with your Nvidia purchase/support, avoid the headache and exercise your option to buy another competitor.

 

[Remij]

Speaking of profiles does anyone remember the quack3.exe scandal ?
where nvidia were detecting quake3 and lowering image quality to inflate benchmark scores caught when someone hex edited quake3.exe and replaced all instances of quake3 with quack3 and the framerates dropped (ati also did something similar) getting caught didnt stop nvidia from cheating they just used more sophisticated methods of detecting games and benchmarks like fingerprinting the camera path in 3dmark and adding static clip planes to the scene.
no ones suggesting nvidia are still doing that but then again is anyone testing for it?

I remember it. Now games purposefully lower image quality to inflate performance for marketing reasons..lmao. So who cares what vendors do at this point? Every vender should "cheat" every single step of the way... as long as visual quality isn't impacted and the game works/doesn't break. There's no such thing as "the right code path" there's code that works better here, and code that works better there.. and if a vendor can manipulate the code to their advantage given their architecture.. then they should probably do it.

As for whether anyone is testing for it.. well, I suppose from time to time reviewers should test between AMD and Nvidia and compare image quality and rendering quality, just to make sure there isn't anything too egregious going on.

 

[Albuquerque]

I found the conversation about driver optimizations, how ethical they might be, and how reasonable they should be expected to be, as an interesting topic worthy of not being buried in the NVIDIA Driver Announcement thread.

So, enjoy the new space to continue the convo

 

[arandomguy]

Speaking of profiles does anyone remember the quack3.exe scandal ?
where nvidia were detecting quake3 and lowering image quality to inflate benchmark scores caught when someone hex edited quake3.exe and replaced all instances of quake3 with quack3 and the framerates dropped (ati also did something similar) getting caught didnt stop nvidia from cheating they just used more sophisticated methods of detecting games and benchmarks like fingerprinting the camera path in 3dmark and adding static clip planes to the scene.
no ones suggesting nvidia are still doing that but then again is anyone testing for it?

I've seen this being rebrought up a few times but it's misattributed to Nvidia. Quack/Quake was ATI.

How ATI's drivers 'optimize' Quake III

techreport.com

What you might be thinking of with Nvidia is possibly the controversey involving shader replacement and 3D Mark (which was much more relevant in the 2000s) which wasn't unique to Nvidia (ATI also did the same). But I guess revsisting this would involve rehashing I guess debates around shader replacement, benchmark/specific software optimizations, and how much partnership developers should have with IHVs.

 

[arandomguy]

As for whether anyone is testing for it.. well, I suppose from time to time reviewers should test between AMD and Nvidia and compare image quality and rendering quality, just to make sure there isn't anything too egregious going on.

If you look at most tech reviewers they want to avoid the controvsey of adding in subjective user experience in favor of just numbers they can use as a shield for objectivity.

I've repeated this feeling before, but modern "reviewers" aren't really doing product reviews for GPUs but just putting out benchmark numbers for the most part. Very few of them actually evaluate or give opinions with a focus on user experience.

 

[digitalwanderer]

I've seen this being rebrought up a few times but it's misattributed to Nvidia. Quack/Quake was ATI.

How ATI's drivers 'optimize' Quake III

techreport.com

What you might be thinking of with Nvidia is possibly the controversey involving shader replacement and 3D Mark (which was much more relevant in the 2000s) which wasn't unique to Nvidia (ATI also did the same). But I guess revsisting this would involve rehashing I guess debates around shader replacement, benchmark/specific software optimizations, and how much partnership developers should have with IHVs.

I remembered it as nVidia cheating too, thanks for the correction.

 

[Remij]

I found the conversation about driver optimizations, how ethical they might be, and how reasonable they should be expected to be, as an interesting topic worthy of not being buried in the NVIDIA Driver Announcement thread.

So, enjoy the new space to continue the convo

Thanks. Yea, it is kind of an interesting topic.. I haven't actually thought about it in ages. A wave of nostalgia hit me when I read his post

My thought is that basically, as long as the vendor are going their own optimizations/cheats driver side.. I could care less as long as the experience is as it should be and nothing breaks. I mean, people take things created by others and modify them to work in different ways all the time.. why not a GPU vendor? As long as nothing breaks and quality is essentially the same.. I'd very much like for them to make things run as fast as possible. I actually expect it.