ATI cannot make working video drivers

[audiofreak]

I don't need glasses and the colored area is the only area you should be concerned with. This same AF test has been used for years on most tech sites when reviewing video cards and never once that I can recall has a reviewer ever mentioned being concerned with the gray area. The colored circle is the benchmark sample. If you can find information from a credible tech site that says the gray area is important as well, please post it. The colored area is what determines the quality of the filter.

"GRAY AREA" IS THE TEXTURE that the filter is being applied to!

There should be a mandatory intelligence test before granting people Internet access, and it should require an IQ of at least 120. Forums would then be a much better place without all those two-digit-IQ trolls.

EDIT:

I got an advice via PM to stay calm -- I appreciate it, I have toned my post down a bit, and I am doing my best not to lose temper, but it is really hard to tolerate people who know nothing about the subject yet feel invited to comment with authority.

Split screen render of anisotropic filtering as a final proof:

Scale 1.0x (default):
http://img19.imageshack.us/img19/681...ghqualityv.png

Scale 0.5x (banding even more obvious):
http://www.abload.de/img/5870tmuvsalu0.5xsduj.png

For uninitiated, left side is rendered by ATI HD 5870, right side is how it should look (rendered by reference rasterizer).

Here is the program if you want to test it yourself:
http://www.3dcenter.org/3dtools/filter-tester

This epic anisotropic filtering fail is present since HD 5xxx series launch, that is obvious from initial filter test screenshots which you can find in numerous reviews.

If it was software problem ATI would fix it silently, and I would not be able to get the same result with the latest ATI drivers.

Think about it, if ATI could not fix this in software for a whole year, then it must be a hardware limitation. What is extremely disappointing for me is that it wasn't present in HD 4xxx series.

[Carfax]

Audiofreak, you're wasting your time dude.. You know the old saying,"You can lead a horse to water but you can't make him drink."

Well it applies here.

Some people just don't want to admit that ATI (now AMD) has inferior texture filtering to Nvidia.. It's always been like that, since Nvidia came out with the G80.

Anyway, the price for apathy is that AMD will continue to short it's customers in that department.

We'll see in a few weeks whether the 6xxx series has the same problem..

[audiofreak]

Audiofreak, you're wasting your time dude.. You know the old saying,"You can lead a horse to water but you can't make him drink."

Well it applies here.

Some people just don't want to admit that ATI (now AMD) has inferior texture filtering to Nvidia.. It's always been like that, since Nvidia came out with the G80.

Anyway, the price for apathy is that AMD will continue to short it's customers in that department.

I know, sigh... I am not doing this because of those "horses". I am doing this so that other enthusiasts can make an educated choice if they care about image quality like I do.

We'll see in a few weeks whether the 6xxx series has the same problem..

Well, TwL of Guru3D (the guy who makes modified Catalyst drivers) thinks that the anisotropic filtering problem may be with drivers. If that is true, then it won't work properly on HD 6xxx as well.

On the other hand, if the problem is in hardware (some people speculate it may have to do with too small of a texture cache size), then HD 6xxx might have a fix for that or might not.

[Carfax]

I know, sigh... I am not doing this because of those "horses". I am doing this so that other enthusiasts can make an educated choice if they care about image quality like I do.

This I can understand. I started a big thread about this same issue about a week ago, and it caused some consternation

Anyway, I think people are finally catching on. I've noticed quite a few complaints on various forums from AMD owners complaining about poor texture filtering on their cards, and many are either returning or selling them and buying Nvidia cards instead.

Case in point.

Now all we need is a big site like Anandtech to focus on this issue when the 6xxx series is released. It will be redemption, since Anandtech was the site most responsible for perpetuating the idea that AMD had "perfected" it's anisotropic filtering.

Well, TwL of Guru3D (the guy who makes modified Catalyst drivers) thinks that the anisotropic filtering problem may be with drivers. If that is true, then it won't work properly on HD 6xxx as well.

On the other hand, if the problem is in hardware (some people speculate it may have to do with too small of a texture cache size), then HD 6xxx might have a fix for that or might not.

Yeah I've heard similar ideas as well. Seems like it would be too much of a gross oversight for it to be a hardware issue, but you never know..

[SoulsCollective]

Carfax, I suggest you read the fourth page of your Anandtech linked thread more closely. This problem, while still under investigation, is more probably related to incorrect driver behaviour or game bugs than outright "bad" filtering.

[Carfax]

Carfax, I suggest you read the fourth page of your Anandtech linked thread more closely. This problem, while still under investigation, is more probably related to incorrect driver behaviour or game bugs than outright "bad" filtering.

Depends on who's explanation you believe..

BFG10K who has researched this issue perhaps more than anyone said on the fourth page:

These are not isolated cases; they happen in a large range of games, even in games that explicitly set trilinear like Doom 3.

Furthermore, ATi's 4xxx series does not show banding in the same situations.

Source

I'm not sure it's a driver bug, because AMD has never provided a way to disable any of the optimizations (which I find suspicious), and the people that have reported this problem to them, have all been ignored.

And it certainly isn't a game bug, since far too many games seem to be affected.

[audiofreak]

It is blindingly obvious that it is not a game bug, because it can be reproduced accurately with AF tester application.

Unfortunately, driver bug cannot be ruled out easily, because it is possible for a bug to affect only one product family -- drivers sometimes do different things with hardware depending on PCI device ID.

What I personally find very hard to swallow is the idea that this is the result of incompetence.

It might be a performance "optimization" or plainly speaking a cheat, because filtering is of much lower quality than expected and it must be computationaly cheaper as well.

The question is, if you REALLY have 2 TFLOPS of performance (like advertised for HD 5850), why bother with such "optimizations" at all?

That leaves us with a hardware problem as the last viable option -- you either don't have the advertised performance and you are cheating in the driver, or you have a design oversight and you are ignoring all complaints and hoping reviewers do not notice until your next hardware refresh is ready.

In my opinion it would be less damaging for ATI if this was indeed a hardware problem, if they admitted it, and fixed it in 6xxx series.

The problem that remains is that regardless of the cause and justification, I still feel cheated -- I paid for "best IQ", not for this crap.

[Blkout]

There's two or three of you guys in this thread pissing and moaning about ATI's drivers acting as if you have so much infinite wisdom. Makes me wonder why ATI isn't employing you guys to work for them since you know so damn much. Apparently ATI's engineers and all of the major tech sites are clearly not as intelligent as you guys. I guess they didn't take audiofreaks IQ test before becoming engineers and professional tech reviewers, good thing we have a few of you guys to keep everyone straight.

[Blkout]

"GRAY AREA" IS THE TEXTURE that the filter is being applied to!

And again, do you have a credible link showing that particular AF test and what part of the image is actually being tested since everyone has been looking at the rainbow circles for as far back as I can recall. As I stated before, post a link to a credible source showing that the gray area is what's being filtered and not the circles. I won't hold my breath. Post a link or give up your ridiculous argument.

[SoulsCollective]

Blkout, the whole test is relevant. The whole thing. The pretty flower in the middle illustrates one aspect, which is angle-dependency. The grey lines illustrate problems with banding or transitions between complex, busy textures. Doing well on one part is great, but doesn't excuse doing poorly on the other.

[rintamarotta]

So your solution is that we all buy two full blown PCs, one with NVIDIA and one with ATI card instead of having them fix things that are broken?

Brilliant mind... so much wasted potential in playing games on two PCs.

I got 3 full blown gaming pc's at the moment, HTPC and those in sig's and theyre all in use by me and my wife or then if its not in use, its crunching at world community grid and gpugrid, none of the power goes wasted i have in here.

My solution works for me, it might not work for someone else.

[audiofreak]

And again, do you have a credible link showing that particular AF test and what part of the image is actually being tested since everyone has been looking at the rainbow circles for as far back as I can recall. As I stated before, post a link to a credible source showing that the gray area is what's being filtered and not the circles. I won't hold my breath. Post a link or give up your ridiculous argument.

In order to be able to understand the test and significance of its components, you would need to have an in-depth knowledge of the following concepts:

1. Texture mapping
2. Mipmaps
3. Texture filtering

And finally:

4. Anisotropic filtering

From your posts so far it is obvious that you do not possess such knowledge.

Going forward, credibility of the source is determined by the level of knowledge of the person who is judging the source credibility.

In other words -- if you knew what I was talking about and how things (should) work in 3D graphics, it would automatically make sense to you, and you would not have to ask for proof other than the one provided by your own eyes.

[audiofreak]

I have finally deciphered what those harsh transitions mean -- ATI is using bilinear filtering between mipmap stages instead of trilinear in their anisotropic filter.

In other words -- they are cheating.

Reference:

Trilinear filtering is a remedy to a common artifact seen in mipmapped bilinearly filtered images: an abrupt and very noticeable change in quality at boundaries where the renderer switches from one mipmap level to the next. Trilinear filtering solves this by doing a texture lookup and bilinear filtering on the two closest mipmap levels (one higher and one lower quality), and then linearly interpolating the results. This results in a smooth degradation of texture quality as distance from the viewer increases, rather than a series of sudden drops.

Underlined part is what we are seeing in those ATI HD 5xxx sample images.

Please note that anisotropic filtering itself only takes into account the viewing angle (basically it corrects for the perspective), but it still consists of trilinear filtering or in ATI's case bilinear because they are cheating.

I am really surprised that they are getting away with it for a whole year now, it really shows how incompetent (or corrupted, can't decide which one is worse) are all those IT review sites that I used to trust to provide me with an honest review.

Regardless of how happy are the armies of ATI fanboys with their uber cards and drivers, it seems that Mr. Terry Makedon is still not satisfied with the current situation -- he has even hired an expert to further improve what ATI is already doing best to their customers:

[mreld3r]

lotta good info here

[audiofreak]

We are in the news. In other words, what we have been saying about filtering is true. There you have it, fanboys.

[Falkentyne]

I wonder what the fanboys who were bashing your posts above, will say, now that you've been proven true.

Apology? Probably not..way too many big egos...but I've been wrong before...

[PaganII]

Apology? Yeah I always base my Vcard purchases by what Youtube looks like ha ha.

[Final8ty]

I wonder what the fanboys who were bashing your posts above, will say, now that you've been proven true.

Apology? Probably not..way too many big egos...but I've been wrong before...

A few reasons why i think most will not give him one.

1)Most people didn't care in the first place & it was only Track mania on stadium that it was noticeable & that even i noticed it way back from when i was playing that game with my ati 1800xt.
Most didn't get upset then so why would they now.


2) The way the message is brought across at times can be more important than the message its self depending on what's its about & he failed big time.

[Kingcarcas]

Not only that but their cards don't run GPUGrid.

[audiofreak]

Let me say upfront -- I do not expect any apologies. I wrote an article and made a thread so that people who want to know the truth can find it.

The way the message is brought across at times can be more important than the message its self depending on what's its about

If presentation is more important to you than the truth, then go and watch Fox News.

& he failed big time.

The only failure is of those who did not understand

[fantomx]

As the series 68xx arived, it look's like the problem with AF was on hardware level, not with software (drivers)...

[Final8ty]

Let me say upfront -- I do not expect any apologies. I wrote an article and made a thread so that people who want to know the truth can find it.



If presentation is more important to you than the truth, then go and watch Fox News.



The only failure is of those who did not understand

I as i said it depends on the importance of the subject & the subject is not important enough for that attitude to be acceptable & people tend to ignore a raving lunatic & people already knew about the issue but you were not happy that they didn't feel the way about it like you did & you through a rant tantrum, thus you don't get apologies.

[XSAlliN]

Audiofreak, you're wasting your time dude.. You know the old saying,"You can lead a horse to water but you can't make him drink."

Well it applies here.

Some people just don't want to admit that ATI (now AMD) has inferior texture filtering to Nvidia.. It's always been like that, since Nvidia came out with the G80.

Anyway, the price for apathy is that AMD will continue to short it's customers in that department.

We'll see in a few weeks whether the 6xxx series has the same problem..

Same goes for nVidia and same goes for you... Not that it was a surprise, considering you signature aka 3 x nVidia cards, probably just an nVidia fanboy bashing an ATi thread.

Owning an ATi card - I'm aware it has problems... - which started being obvious since Catalyst and so on... so no, I'm not a fanboy - you don't see me in nVidia dedicated section bashing nVidia for doing similar things, unless I'll become an nVidia card owner - and being on neutral ground "that is possible" - if I'll feel the need to invest in a future VGA (yet currently nothing impressive - on both sides).

[Carfax]

Great spiel

AMD already admitted though that the 5000 series has a hardware issue when it comes to texture filtering, so anything else is just lip service at this point..

[Final8ty]

Great spiel

AMD already admitted though that the 5000 series has a hardware issue when it comes to texture filtering, so anything else is just lip service at this point..

I will get back to you on this as i play TMUF allot & with the hacked drivers to unlock the features that are only officially enabled for the 6000 series in on 5970s, the issues of texture filtering seems to have gone by setting the new Texture Filtering option to high & disabling Surface Format Optimization.

Have you got a link that ATI/AMD says its a hardware problem.