ATI Responds to Trilinear cheating allegations

[vasileiou]

The Official response comes rom Rage3d

----------------------------------------------------------------
There has been a lot of discussion about our trilinear filtering algorithms recently.

The objective of trilinear filtering is to make transitions between mipmap levels as near to invisible as possible. As long as this is achieved, there is no "right" or "wrong" way to implement the filtering.

We have added intelligence to our filtering algorithm to increase performance without affecting image quality. As some people have discovered, it is possible to show differences between our filtering implementations for the RADEON 9800XT and RADEON X800. However, these differences can only be seen by subtracting before and after screenshots and amplifying the result. No-one has claimed that the differences make one
implementation "better" than another.

Our algorithm for image analysis-based texture filtering techniques is patent-pending. It works by determining how different one mipmap level is from the next and then applying the appropriate level of filtering. It only applies this optimization to the typical case - specifically, where the mipmaps are generated using box filtering. Atypical situations, where each mipmap could differ significantly from the previous level, receive no optimizations. This includes extreme cases such as colored mipmap levels, which is why tests based on color mipmap levels show different results. Just to be explicit: there is no application detection going on; this just illustrates the sophistication of the algorithm.

We encourage users to experiment with moving the texture preference slider from "Quality" towards "Performance" - you will see huge performance gains with no effect on image quality until the very end, and even then, the effect is hardly noticeable. We are confident that we give gamers the best image quality at every performance level.

Microsoft does set some standards for texture filtering and the company's WHQL process includes extensive image quality tests for trilinear filtering and mipmapping. CATALYST passes all these tests - and without application detection, which could be used if you wanted to get a lower-quality algorithm go undetected through the tests.

Finally, ATI takes image quality extremely seriously and we are confident that we set the bar for the whole industry. We don't undertake changes to our filtering algorithms lightly, and perform considerable on-line and off-line image analysis before implementing changes. This algorithm has been in public use for over a year in our RADEON 9600 series products, and we have not received any adverse comments on image quality in that time. If anyone does find any quality degradation as a result of this algorithm, they are invited to report it to ATI. If there is a problem, we will fix it.

[bloodbob]

ATI is currently out of the woods for disabling with colour mip-maps I"m currently getting some 9600/x800 users to do a test to Verify that ATI is telling the truth about the coloured mip-maps excuse.

[saturnotaku]

I wonder how much overtime they had to pay their spin doctors for that bit of claptrap.

If anyone does find any quality degradation as a result of this algorithm, they are invited to report it to ATI. If there is a problem, we will fix it.

ATI has seen the allegations (otherwise they wouldn't have put their PR team to work on crafting a response). Would it be too much trouble for them to test this out themselves?

[smeagol]

I wonder how much overtime they had to pay their spin doctors for that bit of claptrap. ATI has seen the allegations (otherwise they wouldn't have put their PR team to work on crafting a response). Would it be too much trouble for them to test this out themselves?

Ati stated they have investigated the imapct of their AF algorithm before they implemented it.

Finally, ATI takes image quality extremely seriously and we are confident that we set the bar for the whole industry. We don't undertake changes to our filtering algorithms lightly, and perform considerable on-line and off-line image analysis before implementing changes. This algorithm has been in public use for over a year in our RADEON 9600 series products, and we have not received any adverse comments on image quality in that time.

[Socos]

I wonder how much overtime they had to pay their spin doctors for that bit of claptrap.



ATI has seen the allegations (otherwise they wouldn't have put their PR team to work on crafting a response). Would it be too much trouble for them to test this out themselves?

I believe they have. It was not found out because IQ was degraded, I believe it was discovered because someone noticed that when color'ed mipmaps were turned on performance went down.

I think with anything you should be able to turn it off when it is not wanted, so I do disagree slightly with ATI's approach, however, at the same time if I cannot see an IQ drop and get increased performance then great!!

[gordon151]

I know, there should be an option to turn it off even if it doesn't improve image quality by doing so. Also they should have told us a year ago about this method when they started using it on the RV3x0 series.

[Ruined]

In actuality, I don't really mind the optimization - though one problem with it is that reviewers turned on "full trilinear" & highest quality with Nvidia drivers and thought they were getting the same with ATI cards, but they weren't. The quality of ATI's optimization seems to lie somewhere in between Nvidia's Brilinear and Full Quality trilinear. The primary problem with it is that it cannot be disabled even at highest quality settings, and reviewers don't really have an equal quality Nvidia filter to compare it to, only a higher and lower quality filter.

This makes me think of is the "apples to apples" comparisons people were doing with the FX series. I think ATI's optimization is fine in terms of IQ (as I think most of Nvidia's FX-seres ones were), but if you are going to give people the impression you are doing full trilinear and then fail to do it when your card is most definitely capable of it and also criticize your competitor for not doing full trilinear, it's no different than when Nvidia was "violating" the whole "apples to apples" deal last year. Yes, this ATI optimization is less aggressive than Nvidia's FX-series Brilinear filter, but it is similar IQ-wise to some of the shader replacements Nvidia did - not mathematically the same but similar output.

The company that was screaming for a year about optimizations and rendering exactly what the app asks for has now implemented an optimization that makes them look hypocritical. The ironic thing is, it took ATI's own optimization to perhaps get people to the point that zooming in at 400x and using image subtraction as was done with this ATI optimization and many Nvidia optimizations in the past might not be an effective way to evaluate an optimization. In other words, I think ATI's "cheating" has caused an evolution in thinking in terms of optimizations and image quality. OTOH, their statement also justifies a large amount of what Nvidia did last year, and contradicts a large amount of what they were saying.

All in all, again I really don't mind about this optimization (though I didn't mind about many of Nvidia's optimizations in the past)... I wonder how will others look back at some of the more mild Nvidia optimizations that did not create "mathematically equivalent" output but similar output, just like this ATI optimization.

[gordon151]

In actuality, I don't really mind the optimization - though one problem with it is that reviewers turned on "full trilinear" & highest quality with Nvidia drivers and thought they were getting the same with ATI cards, but they weren't. The quality of ATI's optimization seems to lie somewhere in between Nvidia's Brilinear and Full Quality trilinear. The primary problem with it is that it cannot be disabled even at highest quality settings, and reviewers don't really have an equal quality Nvidia filter to compare it to, only a higher and lower quality filter.

This makes me think of is the "apples to apples" comparisons people were doing with the FX series. I think ATI's optimization is fine in terms of IQ (as I think most of Nvidia's FX-seres ones were), but if you are going to give people the impression you are doing full trilinear and then fail to do it when your card is most definitely capable of it and also criticize your competitor for not doing full trilinear, it's no different than when Nvidia was "violating" the whole "apples to apples" deal last year. Yes, this ATI optimization is less aggressive than Nvidia's FX-series Brilinear filter, but it is similar IQ-wise to some of the shader replacements Nvidia did - not mathematically the same but similar output.

The company that was screaming for a year about optimizations and rendering exactly what the app asks for has now implemented an optimization that makes them look hypocritical. The ironic thing is, it took ATI's own optimization to perhaps get people to the point that zooming in at 400x and using image subtraction as was done with this ATI optimization and many Nvidia optimizations in the past might not be an effective way to evaluate an optimization. In other words, I think ATI's "cheating" has caused an evolution in thinking in terms of optimizations and image quality. OTOH, their statement also justifies a large amount of what Nvidia did last year, and contradicts a large amount of what they were saying.

All in all, again I really don't mind about this optimization (though I didn't mind about many of Nvidia's optimizations in the past)... I wonder how will others look back at some of the more mild Nvidia optimizations that did not create "mathematically equivalent" output but similar output, just like this ATI optimization.

The problem is ATI's adaptive filtering method seems to be visually similar to full on trilinear while nVidia's brilinear method isn’t (at least as far as my eyes are concerned). ATI's essential claim has been that if it doesn't reduce IQ, then they're ok with it. That isn't enough for a lot of enthusiasts though since nVidia has everyone paranoid about optimizations and by that alone two errors ATI made were to not openly disclose this and not having an option to turn it off. As DaveBaumann has demonstrated, it has existed in the RV3x0 line and went unnoticed for well over a year now despite possibly hundreds of reviews and IQ analyses. I'm just surprised it's being discovered now.

[jimmyjames123]

Actually, some reviewers have been noting that the 6800U has clearer/sharper AF in some instances vs the X800 cards.

http://www.firingsquad.com/hardware/ati_radeon_x800/page9.asp

http://www.tomshardware.com/graphic/20040504/ati-x800-34.html

[Lfctony]

I didn't mind Brilinear, I don't mind this one either, and that because in either case, I couldn't notice. I sure hope though that there is some way to disable it in the the drivers. Too bad you couldn't and still can't in the FX series.

[D.K.Tronics]

It's a shame the ONLY overdrive feature thats working, is ATI's PR Machine.
I can't believe that there are people out there, on various forums (cough DriverATInHeaven, BeyondATI) praising ATI for this PR FUD, and practially throwing their arms up in the air and rejoicing.
I mean, WTF !! ? So now it's ok to cheat ?

If there was an option in their drivers, and EVERY single review stated that they were using this, it wouldn't be quite the problem that it is.
Other Questionable things are

ATI told reviews to make sure NV40 was doing FULL trilinear, while their cards were doing this 'Optimisiation'.

Shipped out cards without the Overdrive feature working, does it still work ?

Shipped out 'Special' hand picked cards that were clocked higher, like Extreme Tech's card which was clocked 50mhz higher on both the core and memory, yet info was left out of reviews. And Overdrive was purposely disabled in the BIOS, but worked in the drivers, as demonstrated by installing a 9800XT in the same PC.
Was this fair ? Benching a higher clocked card, to a normal clocked NV40.

The whole hypocrisy going around the websites, and forums, is just un freaking believable.
Are ATI now allowed to be let off so easily, from now on ?

We all know that if said websites above, found 3 tiny red pixels, in Farcry on NV40, instead of brown, there would be 4 million posts of "Cheaters", "I'm never buying their crap again", and screenshots blown up 400%, followed by 'pointing fingers' style comments, and NVidia hatred campaigns/witchhunts.

In the end, I really don't care what ATI does. It's just that, when it's Red it's dead, and when it's Green, it's GOT to be seen.

[Nv40]

ATI told reviews to make sure NV40 was doing FULL trilinear, while their cards were doing this 'Optimisiation'.


by looking at their guideline optimizations for benchmarking.. yes.
it shows that they are not confident about their performance. Nvidia NV40 cards in most of the game benchmarks were ahead until AF was used in ATI cards.

So now we have...
if you force TRILINEAR throught [aplication preference] and registry hacking ,you get their new "optimization", ATI-tropic filetring.? and if you use [High quality settings] in the control panel you get ATI-tropic filtering in stage0 and plain Bilinear in the textures 1-7 ?


i dont mind Optmizations.. but when they are done in secrecy,for the sole reason of win benchmarks (otherwise full High quality settings will not be hidden) at the expense of others ..its makes pointless any comparison. between ATI and NVidia. ANd gamers and reviewers are thinking they are benchmarking FUll TRILINEAR ,but the truth is other .

in the end REviewers and JOurnalist are part responsible for all this.
ATI and NVidia feedback of the community is the one that push them to increase or improve their IQ and Perfomance. the problem is that ATI has always been allowed to play their tricks with AF. very few sites.. mostly foreign languages critizised them. while Major sites praised them. Nvidia is the only company that implemented FULL TRILINEAR non angle dependent . Brilinear and angle dependent came because of the Free pass most reviewers gave to ATI there. still today ATI offers 5 years old Bilinear AF in their maintream R9200 cards. while the FX5200 offers much better quality.

the diference with NVidia BRilinear it that here is was not done in secret.. any colormipmpas aplications could see it. and they never claimed to do FUll trilinear there .still this optimization exist in FX line ,
and still today they recomment to use brilinear .so people know what they get when they use HIgh quality settings. since everyreviewer is fully aware of it.if someone is not ok with that optimization then they have the choice to by another hardware that offers what they want. so noone is cheated ,you get what what you saw in the reviews.

Now in the NV40 ,Nvidia choose to do the same angle indepent ANiso because thats what most biggest reviewers (we know who they are worship) FAst and "doenst trade too much IQ ". however ATI now goes one step down and offers again Brilinear.. but this time they do it in secret. what is worst is that they claim full trilinear in their slides,(as default option) and point their fingers to Nvidia for not making it default. when the truth is another. and unfortunately here you dont get what you saw in reviews. because if it were not for German reviewers that investigate more ,and saw what was happening behind the scenes this new "optimization" from ATi will never be know that existed.

[Lfctony]

E Vasileiou, perki na eisai kai Kyprios.

[D.K.Tronics]

Are you guys having forum probs ?
Haven't been able to get on them for a while.

Anyway, while browsing around Futuremark's forum, I found this.


ATI Responds to Trilinear cheating allegations
http://homepage.ntlworld.com/susan.poole2/nocheatATi.jpg

Lmao. Well I thought it was funny.

[MikeC]

It's our new server it has a problem we are working on.

[aa004]

Actually, some reviewers have been noting that the 6800U has clearer/sharper AF in some instances vs the X800 cards.

http://www.firingsquad.com/hardware/ati_radeon_x800/page9.asp

http://www.tomshardware.com/graphic/20040504/ati-x800-34.html

Broken record... broken record... broken record... broken record... broken record... broken record... broken record... broken record... broken record...

Even us lurkers can recognize and pay due respect to the new Hellbinder/Chalnoth in training.

[D.K.Tronics]

It's our new server it has a problem we are working on.

You're getting too popular, Mike. :)

[vasileiou]

E Vasileiou, perki na eisai kai Kyprios.

Oxi akribos...Eimai Athenaios :)
------------------------------------

As about ATI cheating in AF(trilinear) or not etc
I think the future will tell

[jimmyjames123]

Broken record... broken record... broken record... broken record... broken record... broken record... broken record... broken record... broken record...

You signed up at NVNews just to post that? That's pretty inane. :screwy:

[Clay]

Just ignore the threadclowns jimmy. ;)

[jbirney]

I mean, WTF !! ? So now it's ok to cheat ?
How are they cheating? Are they lowering IQ to get speed? No they are not. Please show proof if you feel other wise.

Shipped out 'Special' hand picked cards that were clocked higher, like Extreme Tech's card which was clocked 50mhz higher on both the core and memory, yet info was left out of reviews. And Overdrive was purposely disabled in the BIOS, but worked in the drivers, as demonstrated by installing a 9800XT in the same PC.
Was this fair ? Benching a higher clocked card, to a normal clocked NV40.


You must be joking right? If not what do you call the UE that nVidia shipped out? Those were not hand picked as well?

ATI told reviews to make sure NV40 was doing FULL trilinear, while their cards were doing this 'Optimisiation'.

I agree with this part. Not cool. However so far in side by side shots no one can tell the difference. So while shame on ATI for not being as honest, since no IQ is lost is it a big deal?


the diference with NVidia BRilinear it that here is was not done in secret..

BS. They first appeared only when the drivers detected UT2k3.exe, then later showed up in all D3D apps. Its only when we found out about it and pressed NV that they admitted to this optimization. The difference between the two is so far the ATI method does not degrade IQ where as the first version of the NV method did as it was spotted by many reviewers with out the need of mip maps. MikeC even showed us the UT2k3 videos of brilinear in action.

[vasileiou]

Please guys

Do not start acting like Nvidiots and Atidiots

Just do not start a flame war about these....

[aa004]

You signed up at NVNews just to post that?

Edited by MikeC- If you can't post anything of value to the thread go back to lurking because your first 2 posts have been a big disappointment.

[TheTaz]

Well... it's apparent to me that BOTH companies LIE / SPIN. There should be NO optimizations on the "Top Quality" settings, from EITHER company. Period. (Or a way to turn off all IQ optimizations in the control panel).

Time to switch to Matrox or something. :retard: LOL

Ok... so I'm not going to do something stupid like switch to Matrox. But if they had a card that could compete... I'd sure consider it. :argh:

Taz

[Cota]

Please guys

Do not start acting like Nvidiots and Atidiots

Just do not start a flame war about these....

Its NVidiots and fanATIcs :retard: