ATI cheating on anisotropic filtering performance

[Diplo]

ATI launches card. ATI ships cards. They gave us what they said they would give, they made no false claims. I have nothing but the utmost respect for them as a company.

My local supermarket advertises bread and when I went there they sold me bread. They're conveninent for me, but they don't get one iota of mu respect for taking my money to give me a product they advertised.

 

[Joe DeFuria]

ATI launches card. ATI ships cards. They gave us what they said they would give, they made no false claims. I have nothing but the utmost respect for them as a company.

My local supermarket advertises bread and when I went there they sold me bread. They're conveninent for me, but they don't get one iota of mu respect for taking my money to give me a product they advertised.

Forgiving the poor analogy since ATI is not the retailer, but the manufacturer...

So what would happen if they advertised bread, you went there, and they didn't sell you bread?

 

[Diplo]

Forgiving the poor analogy since ATI is not the retailer, but the manufacturer...

Well, actually, they are both (but mostly the latter, I agree).

So what would happen if they advertised bread, you went there, and they didn't sell you bread?

I'd be pissed off, naturally. However, I still don't quite see why some people hold harware companies in such high regard for simply doing what you expect them to do.

 

[Joe DeFuria]

I'd be pissed off, naturally. However, I still don't quite see why some people hold harware companies in such high regard for simply doing what you expect them to do.

I know what you're saying, but the reason why people do hold it in high regard, is because it's so rare to see it in this industry.

 

[Diplo]

I know what you're saying, but the reason why people do hold it in high regard, is because it's so rare to see it in this industry.

Perhaps. Though I do think peoples' expectations and our geekish tendency to polarise issues actually contribute to the way the industry is run. Our insatiable desire to see 'our team' do better than the other actually becomes a driving force and creates this surreal aura that surrounds hardware manufacturers. Perhaps if we didn't worship them so much they wouldn't feel they had to deliver so much at any cost? Perhaps if we didn't elevate them so high they wouldn't crash down with such a bang?

Both ATI and Nvidia (and Matrox, 3DFX etc. before them) have done a lot for the video industry. No one of them is singularly responsible for Everything Great (tm) to happen, and no one of them has a monopoly on dodgy practises. The only thing anyone can say for certain, though, is without competition progress would be stifled. Yet how many people would take great satisfcation in seeing Nvidia/ATI/whoever go bust?

 

[nelg]

ATI launches card. ATI ships cards. They gave us what they said they would give, they made no false claims. I have nothing but the utmost respect for them as a company.

My local supermarket advertises bread and when I went there they sold me bread. They're conveninent for me, but they don't get one iota of mu respect for taking my money to give me a product they advertised.

Yes but is it low k bread? Or is that low carbs

 

[WaltC]

... However, I still don't quite see why some people hold harware companies in such high regard for simply doing what you expect them to do.

The converse makes it more understandable. The only reason for holding companies in low esteem is when they publicly do exactly what you'd never expect them to do, such as lie, cheat...and in the process try to steal their business. That's why the contrast exists, imo. The low-esteem companies hold their customer base in the deepest contempt by assuming such tactics will ever work in the first place, which is what you'd never expect them to do since it must be assumed they are seeking business instead of trying to scare it off...

 

[Bambers]

thats Funny I read at Toms (i need to get the link) That in their AFtester tests the X800 clearly did a better job of covering all the angles than Nvidias new method.

The AF is nearly identical for both. The bad angles are marginally different but overall one doesn't do any more AF than the other.

 

[ERK]

I'm guessing Chalnoth's just shootin' the sh*t, here. *shrug*

 

[Pete]

Chalnoth, WTF is up with this trollish title and post? I expect better from you, and especially now, with the influx of new members and outburst of threads 'n posts.

1. ATi hasn't changed their default behavior, so if you've been paying attn at all for the past year or so, you'd know this is ATi's CP modus operandi.

2. Dave noted that setting AF via the app results in tri across all layers.

3. I'm pretty sure most reviewers enabled AA/AF in-game, where possible. They did so in Far Cry, at least, and that's where the X800 kicks the 6800's ass the hardest.

4. WTF didn't you pick up on Lars' more important point, that he seemed to notice less texture shimmer with the 6800's AF?

If all new games allow control over AA anf AF in-game, I'd let ATi's method CP slide. If ATi will introduce app-specific controls with newer drivers, though, I think their customers will be far better served with full tri on all layers via the drivers.

 

[anaqer]

I'm guessing Chalnoth's just shootin' the sh*t, here. *shrug*

He usually has a bit more style though, and is definitely wiser then to dig up this old story and make it look like it's some earth-shattering scandal - no matter how biased he is. Call me crazy, but I'm trying to give him the benefit of doubt about his intentions... maybe he was drunk, drugged, or somebody hacked his account or something... dunno.


He IS right about the somewhat ridiculous nature of Tom's line though... not because of the stage filtering stuff but because AF is one of the few important features left with no improvements over the R3x0 (AFAIK) - and there really should have been (esp. wrt angle dependancy).

 

[Bouncing Zabaglione Bros.]

Chalnoth has been corrected on this issue many times before. He's been told how ATI's trilinear works, ususally by someone pointing out how Nvidia forces bilinear whether you want it or not, even in the most recent drivers released to sites for a spoiler on the ATi launch.

For Chalnoth to repeat the same old broken record of misleading information again in the face of numerous corrections... well that's just trolling.

 

[Dave Baumann]

Well, I guess this explains why the AF scores in the 61.11's increased in DX:

http://www.techreport.com/reviews/2004q2/radeon-x800/index.x?pg=6

Unfortunately, the 61.11 beta drivers we received tis past Friday night didn't behave as expected. We ticked the checkbox to disable trilinear optimizations, but our image quality tests showed that the driver didn't disable all trilinear optimizations in DirectX games. I did have time to check at least one OpenGL app, and trilinear optimizations were definitely disabled there. We will show you the image quality impact of the 61.11 drivers' odd behavior in the IQ section the review.

 

[thatdude90210]

Well, I guess this explains why the AF scores in the 61.11's increased in DX:
http://www.techreport.com/reviews/2004q2/radeon-x800/index.x?pg=6

That "just say no!" speed optimization should work for AA also.

 

[jimmyjames123]

ATI is still using full bilinear filtering on texture stages other than the base texture [via control panel settings].

Does this imply that the ATI test results will vary depending on whether AF is set via control panel or set via application?

Has anyone here tested how much framerate changes?

 

[karlotta]

Chalnoth has been corrected on this issue many times before. He's been told how ATI's trilinear works, ususally by someone pointing out how Nvidia forces bilinear whether you want it or not, even in the most recent drivers released to sites for a spoiler on the ATi launch.

For Chalnoth to repeat the same old broken record of misleading information again in the face of numerous corrections... well that's just trolling.

Yes its strange, the last big discussion on this was with UT2K3 thread here and he was a big part of that thread and KNOWS FULLWELL how this is done.
maybe he is testing the mods to see if they change the title( pro ATI and all...pifft).

 

[Diplo]

The converse makes it more understandable. The only reason for holding companies in low esteem is when they publicly do exactly what you'd never expect them to do, such as lie, cheat...and in the process try to steal their business.

Maybe I'm cynical, but I reckon most (big) business would lie and cheat to get a bigger market share. Whether it's any better or worse to do it in private or public is debatable.

 

[dksuiko]

Chalnoth:

What happened? We all have our preferences, and yours happens to be nVidia gathering from your past posts, which is fine. I don't believe I've seen you make quite this inflamitory of a post before though. You are a smart guy, and given the number of posts you've made (compared to my measily 300-400), I'd assume that you knew this was the case with ATI drivers since well before the X800XT was launched. Now, I'm not saying that this is a non-event, but it is certainly not a huge revelation like your thread title makes it out to be.

<snip for size>

On the other hand, if the application (or for that matter a user) requests trilinear filtering on all texture stages and doesn't get it, then we have a problem. The same thing is true of rendering shaders, inserting clip planes, or other mischief to try and decieve the user. I don't think the case mentioned in this thread is in the same category.

Nite_Hawk

It is disappointing to see even Chalnoth snoop down to this level and make a post as accusive as this. We don't need any more threads to cause more mud-slinging. Anyway, excellent post Nite_Hawk - saved me the need to make a more detailed post as I agree with every word in yours.

 

[mozmo]

I did suspect ATI might be cutting corners in their AF implementation, because the performance 4x AA on both the ultra and R420 yielded similar performance loses, but with AF the R420 seemed a lot less impacted then the 6800, so obviously using bilinear stages at 16x is gonna give a nice performance boost, ie the card doesn't drop down to effective 8 pipeline performance, since non of these cards can do single cyle trilinear. the of course the extra clock of the R420 helps it nudge ahead further with AF enabled.

 

[Dave Baumann]

For on - take a look at the reviews that use control panel or application AF. Note that out reviews use application where we can, and this is dictated in the Test Setup page.

Also note, that ATI's control panel AF always uses Trilinear on the first texture layer, so thats always on - it only drops to bilinear if multitexturing is used.