Trilinear vs Trylinear vs Brilinear analysis

[Rolf N]

*Sigh*...Some people just have WAY too much time on their hands....

Yeah, but you loves us anyways.

Not referring to people like yourself Dig...I am referring to hardware sites who are using a microscope to find anything they can to create controversy and/or hits for their websites which really have little to no relevance to the average gamer and in most cases the hardcore gamer.

I for one like xbit's analysis, because that's what it is: an analysis. It's the first article on the subject that really tries to take a technical approach. They could just have spewed the usual crap if page impressions were all they wanted. I'm glad they didn't.

Btw, there's an English version available.

At the end of the day, if I see an optimization that is so obvious as to affect the rendering quality of a game or it's intended output while in motion, THEN and ONLY THEN will I consider witch hunts like this one to be of measurable value to consumers. If such a case did exist, then I would jump on the discussion and try to find a solution or reason for such "optimization".

So far, I have not seen anything to convince me that these "investigations" are worthwhile or meaningful and it is getting old and extremely redundant....

The fundamental problem is that even though many people are interested in competitive image quality analysis, the tools used for that purpose so far are inadequate. You said it yourself, you must see the effect in motion, and I agree totally.
What can websites do? About a million screenshots have been posted to the web for every graphics card you can buy. But they are just stills. How can you judge filtering quality by stills? Take Unreal with its negative LOD bias idiocy. Looks brilliantly "sharp" in stills, but in motion it just makes me want to puke.

You'd need to either own the card yourself, or you need high quality, high frame rate video material on the web. The latter is most likely hampered by bandwidth prices more than by the will to do so, or lack of equipment. Anyway, it's just not being done.

I don't know about you (maybe your job naturally enables you to do that), but I have neither the time, money nor motivation to buy every card to just "see how it looks". I want to be informed before I make a purchase. I'll take insightful analysis like the one on xbit/digit-life any day.

 

[dan2097]

Isnt this article COMPLETLY rubbish. I mean isn't the testing methodology completly flawed? ATI claims their new filtering is adaptive9thus allowing it to provide trilinear where required) so ixbts method of testing it is to purposefully fool the algorithm into giving brilinear when in actualality their filtering tests would normally be detected as requiring trilinear.

Unless im completly mistaken, what exactly does that prove in the context of comparing Nvidia and ATIs filtering?

 

[jvd]

I've read all these articles and it makes me laugh still.


I mean come on . There are a few diffrences in a few small spots that you need to blow the picture up 800% or you need to subtract one picture from the other .


My eyes can't do either of these and I'm superman . So what are you ordinary folks going to do ?

In motion on a x800pro there is no diffrence between this card and the 9700pr i've been using for almost 2 years now.

Both look much better than 5800ultra .

Go to a store find a pc with a 9600 in it and ask to run 3dmark (bring it on a cd) and when u leave you will say to yourself. Wow . Looks just like a 9700pro

 

[christoph]

@jvd
if you dont have the technical interest why do you visit a technical forum on a regular basis then?

 

[PatrickL]

@jvd
if you dont have the technical interest why do you visit a technical forum on a regular basis then?

You are on that forum, and if i check your posts since april, i dont think it is the technical aspect your main interest too, unless trying to report all you can against ATI. You were registered on that forum in 2002 and i don't see your obvious technical interest with nvidia behaviour in 2003

 

[jvd]

@jvd
if you dont have the technical interest why do you visit a technical forum on a regular basis then?

because what technical intrest is being discussed here ?


My technical intrest is how this works . But since ati came up with it and aren't trying to have it liscensed by other companys i doubt i will ever know.


Looking at screens to see if one is better than the other is not technical.

None of these discusions are technical they are are simply witch hunts and its starting to get tiring .

None of these articles are show me why this is such a big deal. We get may affect image , may cause that. But for those of us with the cards in question no game is showing these cases .

The person with farcry claiming he can see problems on the wall ? Guess what I don't see problems on the walls . Not on my 9700pro , 9600xt or x800pro .

You ever wonder why these websites aren't throwing up videos ?

Its because the diffrence isn't there. They can only show it by zooming in screen shots or subtracting one image from another .


As I asked can your eyes do either of these things ? Let alone while the images are going by at 60 times a second ?

With the fx series at the start it was very visable both in game and in images . With this it is barely noticable in images and is not noticable in movement.

So really . What is the technical discusion ?

 

[christoph]

You are on that forum, and if i check your posts since april, i dont think it is the technical aspect your main interest too, unless trying to report all you can against ATI. You were registered on that forum in 2002 and i don't see your obvious technical interest with nvidia behaviour in 2003

jeez, i dont see myself bashing ati and praising nvidia in any fanboish way or vice versa nor do i have the interest in doing so. you fail to see my initial point

 

[Quitch]

Trylinear/Brilinear it's all the same as proven, let the benching be on even ground now i say!

What would that even ground be, exactly?

 

[demalion]

Some of the problems I've observed at the site before aren't in evidence in the article, which is an improvement. Also, despite the issues I have with it, there does at least seem to be an outstanding examination aside from them.

My observations:

1. An English translation reveals that the "promising graphs and forumulas" I noticed about the original article do deliver on their promise, and provide an interesting and comprehensive analysis on the LOD formula. This is a good step forward on analyzing hardware in general.
2. They use this determination to propose what seems to be a skewed final evaluation of aliasing and detail (not necessarily in an IHV biased way, however, as the potential error exists for both IHVs) that is flawed not in what it is based on (the LOD examination is excellent, AFAICS), but what it ignores, i.e. that the mip map sampled from is not the only factor determining the resulting detail and aliasing. The methodology behind the samples taken and the interaction with the comprehensive LOD determination investigation seems completely unexamined, and even the "guess" conclusions offered seem to direct away from recognizing that at all by evaluating in their absence.
   Also, of course, if the sampling methodologies involved do not counteract the LOD observations, ignoring them might not be a significant issue. I think IHVs should be providing explanations to address this if their methodologies take this into account.
3. I was disappointed and puzzled, and remain so, that they went through the trouble of offering a program tool in conjunction with the article towards the goal of evaluating the filtering issues, and completely bypassed, AFAICS, any opportunity for something like "moving" across the texture so the user could actually get an in motion evaluation.
   This is an issue of disappointment in how the program tool and its results are represented to exclude important factors, not for going through the trouble of creating a tool (that IHVs should have released, IMO) in the first place.

Overall, the article strikes me as "excellent in what it investigated" and "significantly flawed in what it didn't consider or evaluate, even when offering preliminary conclusions". As most of the article is centered around the first factor, it is still, IMO by far, the best article on the subject I recall.

 

[chavvdarrr]

JFYI you may want to take a look at NEXT article http://www.ixbt.com/video2/nv40-rx800-4.shtml

 

[jvd]

JFYI you may want to take a look at NEXT article http://www.ixbt.com/video2/nv40-rx800-4.shtml

can't read it . But the pics show that the x800 looks just like the 9800 . So whats the problem ?

 

[digitalwanderer]

EDITED BITS: Sorry Dave

 

[chavvdarrr]

JFYI you may want to take a look at NEXT article http://www.ixbt.com/video2/nv40-rx800-4.shtml

can't read it . But the pics show that the x800 looks just like the 9800 . So whats the problem ?

What problem? Just more comparisons... as authors wrote there - "we give you pictirue and our personal opinion... quality is subjective..."

 

[jvd]

JFYI you may want to take a look at NEXT article http://www.ixbt.com/video2/nv40-rx800-4.shtml

can't read it . But the pics show that the x800 looks just like the 9800 . So whats the problem ?

What problem? Just more comparisons... as authors wrote there - "we give you pictirue and our personal opinion... quality is subjective..."

but if you read thier last article there should be problems with the image. This one shows nothing . So i'm asking whats the problem.

 

[chavvdarrr]

these are synthetic tests, and although images are very close, there are differences. And one more part - with games is promised afaik

 

[Sxotty]

I think even ground meant test the 6800 on optimized vs. the x800.

What I want to know, is the 6800 brilinear adaptive as well? People seem to be implying the image looks better than on a 5800, so why would that be if it is not adaptive? Or does the 6800 use a new brilinear implementation. Anyway I don't care too much about most of it.

 

[Hanners]

What I want to know, is the 6800 brilinear adaptive as well? People seem to be implying the image looks better than on a 5800, so why would that be if it is not adaptive?

I think it is better now simply because it is less aggresive than in the initial driver releases it featured in, rather than it being adaptive.

 

[demalion]

jvd, I think you're stuck with having the distortions and spin some people bring to the discussion of this issue in mind. I agree some of the discussion surrounding the issue is circular and concerned with an agenda, but I don't see that as the only way to discuss a concern for image quality with regard to the X800's optimizations (or the 6800's for that matter). Continuing to look for problems on a factual basis is a technical discussion, and treating it the same way as a different type of discussion that happens to discuss the same thing is uncalled for.

...


As far as still screenshots (and no further), the article does seem to show some things:

-

What the example pictures seem to show clearly is that ATI's control panel "quality" option is failing for the texture stage used by the application (looks bilinear in both control panel specified shots).
Also, as expected, that full trilinear is applied to colored mip maps for the X800 with App preferences.

Nothing new there, assuming it was expected and accounted for in the text of the article and for the texture stage used by the testing program.

What is less clear (due to my not understanding Russian) is if the striped texture used is subject to ATI's optimization determination or if it is a case where trilinear is always being turned on because of the way the mip map usage is specified. I'd presume they have tried to make sure that it is managed in such a way that some dynamic aspect of ATI's implementation beyond such cases is tested, but labelling bilinear screenshots as illustrative of the Quality control panel option leaves me with some concerns about control panel malfunctions or some other methodology mistakes, since I don't understand the accompanying text. But, again, they might have explained their intent to highlight (for ATI) the control panel's shortcomings for texture stages beyond texture stage 0.

....

In the worst case, the article might reveal nothing regarding the X800's "trylinear" behavior "in the wild".

Alternatively, presuming that a proper methodology was used, and no other mistakes have been made, it does seem to have confirmed that the mip map transitions are sufficiently anti-aliased as far as can be determined by general still screen shot evaluation, and also to have independently verified another case separate from ATI's test app.


In contrast, nVidia's optimized mode seems to "fail" (in quotes because the results are not at all a failure on nVidia's part as long as the optimization switch in the control panel offers users the ability to choose when to apply it) the still screen shot test...while better than bilinear, the transitions are still prominent.

However, there is the odd case of "performance mode with optimizations off", where there does seem to be reduced blending with colored mip maps, but the lined image mip map transitions are not prominent. Without a translation, I don't see if they discussed this in the context of the testing progam's texture usage, etc.
Since the optimization option is supposed to control bri/tri, and there is no similar success for any other picture where optimizations using reduced mip level blending is indicated, with higher quality specified as well, a methodology mistake or anomalous behavior (like an LOD shift that would be evident in motion as significant aliasing, but have detail in still screen shots) seems likely.

 

[P4-Fan]

I think even ground meant test the 6800 on optimized vs. the x800.

What I want to know, is the 6800 brilinear adaptive as well? People seem to be implying the image looks better than on a 5800, so why would that be if it is not adaptive? Or does the 6800 use a new brilinear implementation. Anyway I don't care too much about most of it.

That's what I'd like to know as well. How do the different optimized settings perform? And what's their IQ like?

Too bad 90% of the people writing/posting here are more interested in pushing an agenda.

 

[Dave Baumann]

What I want to know, is the 6800 brilinear adaptive as well?

No. AFAIK Its fixed at a default API level (although may be altered on a per app basis).