ATI cheating in AF?

[imported_obsidian]

Hmm, so ATI's stance on filtering is the same as NVIDIA's - "If you can't tell the difference who cares." I like that stance (and have for a while). All the people claiming "I WANT TRUE TRILINEAR FILTERING" need to get with the program. They obviously care more about what number they can place on their system rather than how well it plays a game.

 

[Megatomic]

Originally posted by: Dean
Geesh. You guys are acting like the ATI implementation of Brilinear is the Devil's doing when in fact Nvidia has been going it as well for the past two years(and globally at that).

So it was ok to get infuriated and call for jihad when the offender is/was NV, but when it is ATi it's ok to sweep it under the rug as "that's the way it goes these days"? (not directed at you Dean, your comment just summed up what I was seeing in general)

 

[ChkSix]

Dean you're right. But I don't think two wrongs are ok, especially when everyone put ATi up on some sort of pedestal when in reality and over their 15 year (or more) history has made nothing but garbage prior to the 9700 series.

Now, the only thing that truely bugs me, is the hypocrisy of it all. I could care less who is doing cheating today, yesterday or tomorrow for that matter (although I will care when I am getting ready to drop half a grand). What urks me the most are people who conjur up excuses and defend a business ( I hate using that word, cause business ethics more or less are nothing but back stabbing to get ahead while making huge sums of money) that could care less if you died tomorrow. Stock holders are the exception, especially those with large stakes in a company.

Nvidia has been guilty of it, and now ATi is practicing the same horrible methods Nvidia employed to stay competitive with it's NV3x series. I mean even up to the latest generation of cards and Far Cry for instance, the less intelligent among us (not on this forum, just making a general statement) were saying Nvidia was optimizing their code for Far Cry and it can be proven through a renaming of an .exe file, which drops performance. Now, when the tables are turned and they are facing the beat of their own music, they make feable attempts to bob and weave through a minefield..when they shouldn't being doing that because it makes them look ten times as dumb as they might have looked before.

It's like me saying, killing is wrong, I condem it, and then two weeks from now you catch me batting some poor soul's brain into oblivion. When you confront me and say "dude wtf are you doing, I thought you condemned such behavior" I turn around and say to you "well I wasn't killing him, I was just practicing my swing". Would your image of me change? Would you be able to see through my selfish and hypocritical stance? I would sure hope so!

 

[Genx87]

Well said

 

[Ackmed]

Originally posted by: Bar81
Because they're not doing trilinear when they claim they are. All they're doing is dressing it up to sound like an accomplishment when all it is is a cheat to squeeze more performance out of card when that performance isn't there with true trilinear. Also, if this wasn't a cheat then they would give us a brilinear setting and let *us* decided what looks good and what doesn't. If this was nVidia, there would be an outcry en masse.

PS there's no better proof than when a company admits its cheating like ATI has done here.

How did you get they admited to cheating? Did you even read? Have you read the B3D forums, where they did a comparison vs other cards, and they look the same?

 

[ChkSix]

Looking the same doesn't mean they are doing the same thing, especially when one considers the all too important FPS meter.

A quick example:

1. 6800U doing full trilinear and getting 40fps at 1600x1200 4AA/8Anio
2: X800 doing 'brilinear' and getting 45fps at 1600x1200 4AA/8Anio

And now the kicker (the all important FPS meter)

3. 6800U doing full trilinear and getting 40fps at 1600x1200 4AA/8Anio
4: X800 doing full trilinear' and getting 25fps at 1600x1200 4AA/8Anio

Image can look the same, performance is quite different. Would make a huge difference to review sites as well as sway potential buyers.

 

[Dean]

Originally posted by: ChkSix
Dean you're right. But I don't think two wrongs are ok, especially when everyone put ATi up on some sort of pedestal when in reality and over their 15 year (or more) history has made nothing but garbage prior to the 9700 series.

Now, the only thing that truely bugs me, is the hypocrisy of it all. I could care less who is doing cheating today, yesterday or tomorrow for that matter (although I will care when I am getting ready to drop half a grand). What urks me the most are people who conjur up excuses and defend a business ( I hate using that word, cause business ethics more or less are nothing but back stabbing to get ahead while making huge sums of money) that could care less if you died tomorrow. Stock holders are the exception, especially those with large stakes in a company.

Nvidia has been guilty of it, and now ATi is practicing the same horrible methods Nvidia employed to stay competitive with it's NV3x series. I mean even up to the latest generation of cards and Far Cry for instance, the less intelligent among us (not on this forum, just making a general statement) were saying Nvidia was optimizing their code for Far Cry and it can be proven through a renaming of an .exe file, which drops performance. Now, when the tables are turned and they are facing the beat of their own music, they make feable attempts to bob and weave through a minefield..when they shouldn't being doing that because it makes them look ten times as dumb as they might have looked before.

It's like me saying, killing is wrong, I condem it, and then two weeks from now you catch me batting some poor soul's brain into oblivion. When you confront me and say "dude wtf are you doing, I thought you condemned such behavior" I turn around and say to you "well I wasn't killing him, I was just practicing my swing". Would your image of me change? Would you be able to see through my selfish and hypocritical stance? I would sure hope so!

I agree with everything you said. My disgust is aimed towards ATI and Nvidia. Both of them are optimizing and i'm afraid its only going to get worse. I do not know if ATI's optimization was a direct reaction to Nvidia's, but I would hope at least they(ATI) would stick to the high road. Do I feel that ATI's approach to optimizing is as bad as Nvidia's yet? No i don't. But it will not take long to catch up. The only losers will be us.

 

[Dean]

Originally posted by: ChkSix
Looking the same doesn't mean they are doing the same thing, especially when one considers the all too important FPS meter.

A quick example:

1. 6800U doing full trilinear and getting 40fps at 1600x1200 4AA/8Anio
2: X800 doing 'brilinear' and getting 45fps at 1600x1200 4AA/8Anio

And now the kicker (the all important FPS meter)

3. 6800U doing full trilinear and getting 40fps at 1600x1200 4AA/8Anio
4: X800 doing full trilinear' and getting 25fps at 1600x1200 4AA/8Anio

Image can look the same, performance is quite different. Would make a huge difference to review sites as well as sway potential buyers.

I do not think you will see Nvidia doing true trilinear anytime soon. It;s too big of a performance hit still for both companies. ATI never gave true trilinear in the r300 class either and only applied it to the first texture stage. When it comes to trilinear both companies have to make trade-offs or suffer a rather extreme performance hit.

 

[ChkSix]

Exactly Dean. We will be the losers. But one thing I am thankful for, is the open mindedness that exists here and not in any other forums I've participated in. Honestly. At least we can all converse intelligently and constructively without getting silly (silly is great sometimes..hehehe). Elsewhere (and where I have been) most of the responses were denial, immature and downright malicious. You guys here are a step above and beyond that, and although I have been a frequent reader of Anandtech, I regret not joing up to these forums sooner.

Edit: Maybe not Dean, or maybe this is an opportunity for them to prove that they can and do it better. ( Response to Nvidia trilinear statement you made ) I am sure the Nvidia gurus are keeping a close eye on the recent ATi uncovering and working hard to use this in their favor. That's what big business, where finding and using a weakness found in your competior can benefit you financially, do.

But you're right, neither have given true trilinear as of yet.

Mike

 

[Bar81]

OH THE LIES!!! Welcome to the end of the discussion:

Exhibit 1

Exhibit 2

 

[ChkSix]

I love the second one. "Trilinear on by default" Dude I almost spit up my coffee.

 

[Dean]

Originally posted by: ChkSix
Exactly Dean. We will be the losers. But one thing I am thankful for, is the open mindedness that exists here and not in any other forums I've participated in. Honestly. At least we can all converse intelligently and constructively without getting silly (silly is great sometimes..hehehe). Elsewhere (and where I have been) most of the responses were denial, immature and downright malicious. You guys here are a step above and beyond that, and although I have been a frequent reader of Anandtech, I regret not joing up to these forums sooner.

Edit: Maybe not Dean, or maybe this is an opportunity for them to prove that they can and do it better. ( Response to Nvidia trilinear statement you made ) I am sure the Nvidia gurus are keeping a close eye on the recent ATi uncovering and working hard to use this in their favor. That's what big business, where finding and using a weakness found in your competior can benefit you financially, do.

Mike

I think you will see both companies implement some sort of balance. They will give near trilinear quality without a huge penalty. What worries me is that the new r420 is more in line with the 9600's ability with trilinear and not as good as the approach that the r300 used. What it all adds up to is the r420's LOD will be of lesser quality that the r300's.

 

[Dean]

Originally posted by: ChkSix
I love the second one. "Trilinear on by default" Dude I almost spit up my coffee.

From what I read at Beyond 3d that statement is true, but it also fails to say is that ATI can alter that setting on the fly depending on the texture!!hehe Which will problably mean its off alot more often than its on

 

[ChkSix]

Hehehe.

I am sure Fred down at the publications facility of ATi is being led out back to get a good beating right now.

"Damn Fred, how DARE you make a typo and leave out that we can change it on the fly!" You want 3 or 4 lumps with your coffee?! Pow Bang Zoom!

 

[Dean]

My take on this and all the other graphic company fiasco's of late is never take their word for it.

When PowerVR come out with their version 5 it will be the Ultimate cheat i guess. It's HSR ability will not have to render anything near what ATI or Nvidia has to do. Will that make PowerVR approach the Ultimate cheat? or the Ultimate Optimization? The answer is cloudy and every fanboy will have his own opinion

 

[ChkSix]

True that bro.

 

[Genx87]

- Lower filtering (brilinear)

Can you blame them? Since the Cat 3.4s ATI has been doing the same thing.

- 4AA that looks like a 9800s 2AA on most angles

Fixed in 6800

- shader replacement targeted at specific games /benchmarks

Big deal as long as it is mathematically equiv.

- lower shader paths

You mean quality? Depends if the FX was doing it in FP32 or now

- 8AF lowered down to about 2AF in a some benchmarks.

Huh? I think you are confusing the FXs AF with ATIs AF where the ATI version applies 2x AF to everything except 90, 45, 22.5 angles.

 

[Ackmed]

Originally posted by: Bar81
OH THE LIES!!! Welcome to the end of the discussion:

Exhibit 1

Exhibit 2

Once again, games are what matters.

 

[DAPUNISHER]

Corporations are run like war campaigns, period. War is Hell and so is business. Anything that can be used as a tactical advantage and benefits the company is pursued with the great vigor. The enemy is the competitor, if they demonstrate a strategy that is effective then it's adopted, if they show a Chiink in the armor it is exploited.

The hypocracy of condeming one while exempting the other is to be expected. Loyalties are given, tributes paid, aliances formed and desolved. Anyone who contends ATi has somehow never fallen from grace in the course of managing their "campaigns" is naive at best, and in denial at worst.

 

[ChkSix]

Damn Pun, that was pure genius. Awesome post!

Mike

 

[Dean]

Originally posted by: ChkSix
Damn Pun, that was pure genius. Awesome post!

Mike

And its true. Here is another twist. When a company cheats are we to blame the company? Or the competitor for making them do it? hahaha

 

[DAPUNISHER]

Originally posted by: ChkSix
Damn Pun, that was pure genius. Awesome post!

Mike

Thanks brudda :beer: but I can't take credit. It's the principles of Go Rin No Sho the book of five rings by Miyamoto Musashi Japanese Corporations have used it as guide to business strategy for generations and many western companies have embraced it now as well. Once you understand strategy, it is applicable to all areas of life :light:

 

[Matthias99]

This was put up by the Inquirer; it is an official statement from ATI on this:

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.

So... they're not generally doing "full" trilinear filtering with autogenerated mip-maps; they're using some sort of intelligent adaptive algorithm (which, since it's gone unnoticed thus far, must at least be pretty decent). So, while it sounds like they didn't go out and specifically write this to 'detect colored mipmaps', it does have the side effect of giving "full" trilinear on them (and, presumably, any textures with very sharp color transitions between levels), but not on "normal" textures that use autogenerated mipmaps.

However, I personally think there shoud not be *any* silent optimizations like this in the drivers. At the very least, you should be able to disable it, and get "full" trilinear all the time, even if that results in a big performance hit.

 

[DAPUNISHER]

However, I personally think there shoud not be *any* silent optimizations like this in the drivers. At the very least, you should be able to disable it, and get "full" trilinear all the time, even if that results in a big performance hit.

There are reasons that are dubious ones for why they were silent about this, but personally I am all for optimizations that can increase performance without trashing IQ. Perhaps the reason they didn't announce this feature was becuase they were a bit timid given the atmosphere a year ago Besides it might have taken heat off of nV for their optimizations, despite the fact their's did trash IQ in some instances, and you should never interupt the enemy when they are making a mistake.

 

[Matthias99]

Originally posted by: DAPUNISHER

However, I personally think there shoud not be *any* silent optimizations like this in the drivers. At the very least, you should be able to disable it, and get "full" trilinear all the time, even if that results in a big performance hit.

There are reasons that are dubious ones for why they were silent about this, but personally I am all for optimizations that can increase performance without trashing IQ. Perhaps the reason they didn't announce this feature was becuase they were a bit timid given the atmosphere a year ago Besides it might have taken heat off of nV for their optimizations, despite the fact their's did trash IQ in some instances, and you should never interupt the enemy when they are making a mistake.

I'm all for optimizations that increase performance without lowering IQ. However, if you set a graphics card to give you "trilinear" filtering, I would normally expect that to be what some folks over on B3D are calling "naive" trilinear -- that is, complete trilinear filtering on all mip-map levels (which is, indeed, what you get right now if you use colored mip-maps). On the X800, though, you're actually getting some sort of adaptive trilinear, which is not identical to full trilinear in most real-world situations. Now, it may be 99% as good 99% of the time, and if it is, I'm thrilled. But they still should have *said* they were doing this, and given an option to turn it off.