ATI cheating with AF

[Dirty]

http://www.theinquirer.net/?article=15933

WE GOT AN ANONYMOUS mail about ATI optimisation just before everyone started to talk about it last week, but this part of Vienna where I live in was massively hit with Sasser leaving me without or with a deadly slow internet for to long.

We are avid to learn why ATI beta drivers and its ati3duag.dll know what applications are they executing. Wise people that like to look into drivers have found references to these EXE files. You don't have to be wise to know we are talking about games and not about PowerPoint and Winamp here. Here is the short list: ati3duag.dll recognises CT3.exe, pop.exe, RaceDriver.exe, SplinterCell2.exe, SplinterCell.exe.

We will ask ATI for comment in Monday when we would like it to come to us with a clear explanation of what's happening here.

We don’t have any of those EXE files, but in case you are reviewer and have this ATI card please rename them and let me know what happens then. I think that EXEs like FarCry.exe, UT2004.exe would be a much better couice to do some optimisations.

I guess there will be more when ATI decides to comment. µ

[-=DVS=-]

Didn't some german reviews site alread tried renameing FarCry Nvidia the one got big difference in speed not Radeon ?
Whoever got Radeon X800 care to try ??

[bloodbob]

ATI Appears to be doing more or less the same thing Nvidia does with the NV3X cards.

[gordon151]

Dirty, as if it hasn't already been discussed that ATI detects those applications so that they can disable AA. AA doesn't work in any of those applications and people were complaining about the general detection mechanism they used to detect applications that AA doesn't work in. Also the Fartcry tricked had been tried on Radeons and didn't work.

[Evildeus]

Can you provide commentary on what we're looking at with the comparison? Is it better to show more lines/dots/whatever it is in the comparison part?Here it is:
OK, the above link shows some screen grabs of compressonator images - there are X800 XT, 9800 XT and a 5950 for comparison. In all cases bilinear is on the left, trilinear on the right and the middle shows a 400% brightness increase. App is our standard SS:SE texture test using the DX renderer.

[Evildeus]

Looked at what Evildeus linked , but i see no quality difference ingame shot from R9800 vs X800 , only 5950 slightly better on ground btw why no 6800 comparision shots :eek2:[/IMG]
He doesn't have a 6800 at hand ;)

[gordon151]

hmm let's see how the fanboys reply

Well that's one fanboys reply, lets see some more!

[jbirney]

Should be interesting to see ATI's response...and here is what Dave said:


Given that whatever is evident here has been consistent since 9600 was released I would have expected some of the numerous 9600 owners to have complained in the past year if there was some appreciable IQ lowering, however I don't think I've heard it mentioned yet. I'll have to see if I can pull out the old UT2003 maps where its quite evident on the FX hardware.

What I've not yet heard an explaination of, either, is why the Cat3.4 drivers clearly do different things looking at the difference shots between the full texture slider position and the lower slider position and this is reflected in the coloured mip shots - it strikes me based on this that coloured mips are not being "detected".


Then Lar's from THG chimed in with this:
Ok, it seems I need some more coffee this morning

Here are new numbers.

AF forced by driver. 8x AF, Performance (this should result in pure bilinear on all stages)

10x7: 132,85 / 132,84
12x10: 132,12 / 131,08
16x12: 124,68 / 113,85

Now same setting with changed Mipmap Detail Level from H-Quality to High Performance)

10x7: 132,54 / 132,71
12x10: 131,78 / 131,15
16x12: 129,9 / 123,70

The last on with changing Textures from H-Quality to High Performance (Mipmal details back to default)

10x7: 131,22 / 131,92
12x10: 131,18 / 130,69
16x12: 127,84 / 117,41

Well the whole discussion started because the guys at Computerbase found out that the X800 is loosing performance when colored mipmaps are used. The suspicion is that ATI is using a lower filtering by default and switches to full tril only when colored mipmaps are used.

This can´t be the reason since there´s even a performace difference when you´re already using pure bilinear by default!?!? And we have the quote from Epic that there should´nt be a difference at all. Maybe it´s because the lack of coffee this morning but I´m getting a bit confused



So not sure whats going on. But I am a bit tired of both companies trading off IQ for speed just to beat the other....

[gordon151]

That discovery changes things from what was the general presumption yesterday and now I'm just curious as to why on the x800 performance with colored mipmaps drops regardless of settings. Is ATI cheating even at the lowest quality setting?

[saturnotaku]

Is the IQ difference really noticable w/o colored mipmaps?

[jbirney]

Is the IQ difference really noticable w/o colored mipmaps?

Well the article claimes that without the mip maps on it drops to brilinear but so far we dont have any rock solid proof. Dave and Lars have found some inconsistancy in the results.


Is the IQ difference really noticable w/o colored mipmaps?

Lar's had a quote from Epic saying that for UT2k3/4 there should be no change in prefromance with Mip Maps turned on. If there was Epic suggest that Lar's contact ATI and NV about it. Lars showed that both ATI and NV had different numbers when comparing their mip map on to mip map off scores...so who really knows anymore...

[Morrow]

Is the IQ difference really noticable w/o colored mipmaps?

I don't think that the whole affair is about how good ATI's brilinear optimization look but it's rather about how ATI's drivers pretend to do trilinear work but in reality they only do brilinear although full trilinear filtering is forced by the drivers.

These filtering story is about how reviewers try to make benches with both nvidia and ATI cards with best IQ and trilinear but ATI cards are only doing half of the job while pretending being trilinear. I think this is called deceptive advertising and while this would not be anything new for nvidia, it's quite a scandal for ATI if those findings are proved to correct.

In any case the worst part about this is that the catalyst drivers detect coloured mipmaps and enable true trilinear filtering accordingly trying to fool the users.

[hithere]

In any case the worst part about this is that the catalyst drivers detect coloured mipmaps and enable true trilinear filtering accordingly trying to fool the users.

Then how do you explain the performance differences that, assuming what you say is true, shouldn't be there when full bilinear is enabled by default, and you toggle colored mip-maps...YOU KNOW, WHEN THERE'S NO TRILINEAR TO DISABLE. In other words, how do you go lower than the very lowest setting? If ATI is disabling Trilinear for a performance gain, then what are they doing when the user specifies they don't want Trilinear on to begin with? Perhaps they are double-dog-disabling trilinear for when the user doesn't turn it on in the first place? Maybe it's "super-disabled"?

For the games that you can't force Tri on in the control panel, turn it on in-game, and there it is. Is that so hard? Has everyone just lost the ability to find in-game option panels?

[Ruined]

We don't know how much of a performance gain ATI is getting with brilinear, or why across the board scores are lower with mipmaps on than off. We do know from image comparisons, though, that they are doing brilinear with the x800 series - the x800/9800xt AF comparison done at B3D clearly shows the x800 is rendering significantly less trilinear. The real point is that x800 benchmarks at high AF levels can now be regarded as inflated or skewed versus the 6800 on 60.72 drivers. Even if ATI claims this is a "bug" somehow, it still doesn't negate the fact that Nvidia's 6800 on 60.72 is doing full trilinear and x800 is not.

Then how do you explain the performance differences that, assuming what you say is true, shouldn't be there when full bilinear is enabled by default, and you toggle colored mip-maps...YOU KNOW, WHEN THERE'S NO TRILINEAR TO DISABLE. In other words, how do you go lower than the very lowest setting? If ATI is disabling Trilinear for a performance gain, then what are they doing when the user specifies they don't want Trilinear on to begin with? Perhaps they are double-dog-disabling trilinear for when the user doesn't turn it on in the first place? Maybe it's "super-disabled"?

For the games that you can't force Tri on in the control panel, turn it on in-game, and there it is. Is that so hard? Has everyone just lost the ability to find in-game option panels?

[Lfctony]

Then how do you explain the performance differences that, assuming what you say is true, shouldn't be there when full bilinear is enabled by default, and you toggle colored mip-maps...YOU KNOW, WHEN THERE'S NO TRILINEAR TO DISABLE. In other words, how do you go lower than the very lowest setting? If ATI is disabling Trilinear for a performance gain, then what are they doing when the user specifies they don't want Trilinear on to begin with? Perhaps they are double-dog-disabling trilinear for when the user doesn't turn it on in the first place? Maybe it's "super-disabled"?

For the games that you can't force Tri on in the control panel, turn it on in-game, and there it is. Is that so hard? Has everyone just lost the ability to find in-game option panels?

Yes, I agree here. Since brilinear is a mixture of B/Trilinear, when forcing Bilinear, there shouldn't have been a performance drop. Because the driver isn't running Brilinear at all, in order to switch to Full Trilinear. I think its sort of a bug now. Will wait for further information.

[hithere]

Even if ATI claims this is a "bug" somehow, it still doesn't negate the fact that Nvidia's 6800 on 60.72 is doing full trilinear and x800 is not.

...even so, a number of sites doing comparisons in performance initially did so with the "forced brilinear" drivers on the NV side, and it didn't affect the outcome. Further, I believe you can still enable full trilinear on the X800 via in-game option panels for games such as Far Cry. That's ATI's fault for not letting users know how to do so, but when this option is known to users of discussion boards almost a year ahead of time, it ought to be common knowlege for reviewers.

I just hope no one gets me wrong here, I find the whole thing distasteful...I ought to be able to set clearly-labeled options in the control panel of my graphics card, and get those options enabled in-game, or at least have a readily-available and clear explanation of why it isnt' possible with this or that game, and how to get around it. I don't have that with either card, depending on which driver I choose, but have been able to get around it for some time with my 9700. For those who would accuse me of being an ATI fanboy, please know that the 6800 is currently my top choice for next GPU, and will remain so barring some truly awesome performance/IQ advantages in HL2 and Doom3 for the ATI camp...and I doubt that will be the case. I want native stereo support in my drivers, and PS3.0 seems to be working out in Far Cry for the 6800, and the performance advantages the ATI card has aren't all that great, when they are there.

[grimreefer]

in game settings dont overide driver settings. I would see y ati would need to cheat. There card jus isnt as good as the 6800. I would find it hard to believe that nvidia's new architecture would be worse than ati's old 1 in af.

[jbirney]

...inflated or skewed versus the 6800 on 60.72 drivers. Even if ATI claims this is a "bug" somehow, it still doesn't negate the fact that Nvidia's 6800 on 60.72 is doing full trilinear and x800 is not.

I was under the impression that the you still can not get full trilinear on any FX/6800 card yet as the driver switch to shut off brilinear still is not working...Do you have a link that shows it does? Just wondering not trying to flame.

[muzz]

Is this what it has all come down too?
So what if their doin' funny things in the drivers. If you can't see it in the game, then I guess its okay. I just think this is being anal about the whole thing.
nVidia has done worse in the past and so has ATI.

So I think people should just let this slide...

However it is interesting that people can start thousands of topics everywhere else about something as small as 2molex connectors and drag nVidia's name through the dirt. When it's something that affects you directly though (as in what your games look like), it hardly gets as much attention.
ALL the people with the attitude that ATI could do no wrong are nowhere to be found all of a sudden. I bet though is a thread started called "NVIDIA cheating with AF" there'd be 30 replies within 10mins...

Yes NV has taken a beating, that many found justified ( clip planes are justified beating territory, no discussing it).
The thing about this particular issue is the fact that when colored mips are called it go's back into FULL tri mode, which would imply detection, which is BS.
I am having as hard time imagining WHY it would do that, and still hold out any kind of hope that everything is on the up and up. For NV users/fans- I am admitting this looks like a cheat to me with the info we have at this time.

If the image quality is pretty close then cool, but it looks like the mipmaps/full tri was meant to decieve folks into thinking that full tri was always used, when that doesn't seem to be the case.

As stated I will wait till I get an answer regarding this, but things are on shaky ground IMO.

As far as the guy complaing about DaveB @B3d.
As has been stated he did his own testing, and didn't like what he saw either, and HE mentioned this behavior go's back to 3.4 or something.
I swear, 1 post, questioning a well respected 3d site.
Do you work for the NVpr dept. :retard:

[fivefeet8]

I was under the impression that the you still can not get full trilinear on any FX/6800 card yet as the driver switch to shut off brilinear still is not working...Do you have a link that shows it does? Just wondering not trying to flame.

Several previews of the Nv40, before the R420 launch, shows that Full Tri is available in the 60.72 beta forcewares. Anyone can find the articles.

[Lfctony]

The 60.72 work fine, the 61.11 ones are broken though.

[muzz]

Am i the only one who cannot see any of beckman's posts?

Nope I can't see them either.
Haven't been able to read even one of them, and I tried numerous times with different posts!!

[Blacklash]

Look at all the FUD and trash that has been fired at Nvidia since the launch of the 6800. This Ati issue is being discussed on a few message boards but have any hardware sites decided to 'investigate'?

I want the end user to be able to turn off brilinear if they so desire, that goes for BOTH companies. Again if Ati is swapping brilinear for trilinear only when color mips maps/tests are done I would like to see them called on it.

People should stop treating these Companies that take their money like they are their girlfriend. Yes, I know the same dynamic is present :D The fact is we Consumers should not be at each others throats over brand loyalty but rather standing together and holding both Ati/Nvidia to the highest possible standard.

It sucks when Nvidia does sh*t like this and it sucks when Ati does too...

[gordon151]

Even if ATI claims this is a "bug" somehow, it still doesn't negate the fact that Nvidia's 6800 on 60.72 is doing full trilinear and x800 is not.

Correction, in the 60.72's you have the option to turn it off the brilinear filtering optimizations (something absent in the latest set of drivers though). If I'm not mistaken, it was still the default option.

Edit: Malficar, do you really have a 6800-GT?

[ragejg]

Look at all the FUD and trash that has been fired at Nvidia since the launch of the 6800. This Ati issue is being discussed on a few message boards but have any hardware sites decided to 'investigate'?

I want the end user to be able to turn off brilinear if they so desire, that goes for BOTH companies. Again if Ati is swapping brilinear for trilinear only when color mips maps/tests are done I would like to see them called on it.

People should stop treating these Companies that take their money like they are their girlfriend. Yes, I know the same dynamic is present :D The fact is we Consumers should not be at each others throats over brand loyalty but rather standing together and holding both Ati/Nvidia to the highest possible standard.

It sucks when Nvidia does sh*t like this and it sucks when Ati does too...


The way I see it, it is a PRIVILEGE that we, a massed group of people with a similar interest (PC Gamers) are catered to in some regards by two large companies who really make the most of their money with OEM and low end parts...

The developers have the demands and create the pushes forward in technology, and when vidcard company X implements said feature and puts the correct amount of spin on the feature to create a community buzz,which can make or break the success of a new architecture...

The "consumers" mentioned in the above quote make up only a small percentage of sales volume... they are however the loudest demographic, and recieve the most marketing (heck, there's been vidcard ads in Rolling Stone for years...

The companies are doing us gamers a favor. Always. IMO if they screw up once in a while... ok, well you know what, I'm gonna be ok. My house won't burn down because of it. Life will go on.

Because in retrospect, bad drivers or not, bad AF or not, I have been enjoying games courtesy of my video card for years. And the heated discussions, flames and mindless arguments about floating point this and RGMS that... well, IMO those discussions are completely justified... if you're a dev. and you'r ein a dev. forum.

But most of us aren't dev's. We're gamers, that sect of folk who would sleep next to their vidcards if there were antistatic sheets and pillows. :p

I'm ok with R200's AF.

I'm ok with 2xQ AA in some situations.

I'm ok with nv3x's *lower than R3XX* shader performance.

I'm even ok with brilinear.

And though I'm a tad looser in the collar about this stuff than some of you in this thread, I still feel that I got my money's worth with just about every GPU purchase I make. All of these products are amazing in their own respects... I mean c'mon, who would have thought 15 years ago that a piece of PCB like a modern videocard could create amazing scenery found in titles like Silent Hill 3, Doom 3, FarCry etc...?? I still find that fact simply amazing every time I kick into a modern 3d app...

I'll come back to this thread and be a little more on-topic later... I just haven't made a meaty post in a while... so forgive me if what I said above was slightly off key... ;)