A look at IQ at Firingsquad...

[WaltC]

[
Ok, fair enough. What is specifically is wrong with that testing, its methodology, and its results?

I would be skeptical of any results which purport that treating all texture stages with trilinear has no more of an impact on performance than treating a single stage. There may be instances--in cpu limited games, for instance--in which this might appear to be true, no doubt. But it wouldn't be true because there is no performance hit from treating 4 stages as opposed to 1... I'd think the same way about someone telling me that 4x FSAA exacts no more of a performance penalty than 2x FSAA--that might be true in certain games--but not because 4x isn't demanding a lot more overhead...

Hmm, this begs the question as to why NVIDIA is only doing what they are doing in UT2003 and not in all games then, doesn't it?

The other games aren't used nearly so much in web-site benchmarking... nVidia's looking to create dishonest appearances about its product performance, so picking UT2K3 is a specific, targeted choice in that regard. I would exect that if Doom3 were available at present we would see much the same thing in the Dets for that software, for the same reasons.

I agree that filtering things that you cannot see is pointless. What I'm trying to distinguish is how performance is affected.

What would make you think there is no performance difference? Why do you think that, relative to filtering, all IHVs make specific choices as to the texture stages they treat? If you can plainly see the difference in a game in performance between bilinear and trilinear filtering on the visible texture stages that are treated, why would you assume no such performance differential exists for the other texture stages?

Texture stages, texels, etc., are never rendered to the screen in the first place--only pixels are rendered with textures attached--such textures being called texels. But you can't see a texel--only a pixel to which a texel has been attached. If we divide the pre-rendering filtering processing for texels into four stages prior to them being applied to a pixel and rendered to the screen, it should be apparent that the more stages you process, the more work you do, and the lower your performance. This is only "bad" when you filter texture stages which are laid underneath the stage(s) actually visible to the camera (since you can't see them.) To me, as I've stated, the principle is the same as rendering occluded pixels--you simply don't want to render them. The more of them you render, the lower your performance, with no improvement in IQ in the frame.

Are you saying that rTool is not doing what it is alledging that it is doing? (allowing the possibility of trilinear to be used in texture stages where bilinear would normally be used when Quality AF is selected)

I'm saying it's certainly possible it isn't actually affecting the drivers like it thinks it's affecting them in certain stages--but I'm also saying that the bottleneck regarding this kind of performance in games might lie *elsewhere* in the processing pipeline, which would then make it appear in some cases that treating 4 stages is no more degrading to performance than treating a single stage.

 

[demalion]

Which are you practicing with your conclusion? Or does this criteria not apply to yourself? Anyways, I don't think something like abduction should be ruled out as long as it isn't abused to propose certainty.

Does it matter what I am doing?

If that is what I was addressing, well then, yes. And it was what I was addressing. A conversation, yes?

I am asking about your theory.

Well, actually, you're proposing that my theory doesn't successfully describe the flaw in your own statements. Which is what I was using it for. In this conversation.

And I'm answering why it is sufficient for what I was using it for. In this conversation. Where I was using it to point out the problem in your own statements.

Hence, your own statements are relevant.

If there is a flaw in this description, explain what it is If there is not, don't ignore my having established this to later make the same protestation again.

If it is not clear yet, you have already debunked my evidence. I was using the filtered texels as evidence, but if work is the same regardless of how many texels are filtered then my use of evidence is flawed. So I have no evidence to prove or disprove my theory.

Actually, you do have evidence to disprove the statement that the performance difference only occurred with the presence of off angles.

The problem is, I don't see any evidence to prove or disprove yours either.

I can't really control what you do or do not see at any particular moment, since I fail to find self-consistency in evidence in your statements, and you simply continue to refuse discussion about your demonstrating this inconsistency.

I provide evidence. You respond by saying that I didn't, or that it isn't incontrovertible and final and therefore doesn't count. I point out directly to you why such a comment doesn't seem valid. You change your statement to say there is evidence after all, or that you never disputed that I had evidence, but then you go on to ask me to provide evidence again.

I've tried repetition, highlighting (in color and formatting), direct questions, and all have been simply disregarded.

You were saying that in regards to not caring whether the texture aliasing was a result of AF or LOD, as part of "proving" that ATI's AF implementation demonstrated more texture aliasing. This is directly connected to a discussion about AF, yes?

Is LOD the same whether or not AF is enabled?

Hmm? The bias value itself, but not its effects, right? Unless the LOD bias is adjusted along with AF being used, to, for example, increase texture detail without AF processing as I think we discussed, or reduce aliasing while increasing AF processing.

Umm...no. It just isn't determined by only measuring the result you were looking at. How are confusing words like "complete", and my various phrasings of "there is more to look at", with saying this is "indeterminable"?

Then what is it determined by?

By what it actually is, for your usage as far as specifying the methodology exactly. For my usage of addressing your commentary, by the things I listed in response to your queries about "measuring work", as a start.

We need to be clear on this: does ATI's performance hit being reduced depend on the off angles being present, or not?

Yes, I think so.

How am I supposed to deal with such self-contradiction? Upon your providing this reply, should I point out that it is only to reach this point that the evidence you ask for applies, and that I've provided reasoning from this point on already?

Ok, let me clarify what I mean. Yes, ATI's performance hit is lower than NVIDIA's. But it gets even lower when off-angles are present, that is my stipulation.

OK, but this statement, even with your stipulation, supports that "it does more work so it's better image quality" is an invalid precept, which was the point of my conversation.

OK, which theory do you mean? The "theory" I proposed as providing support for my reasoning was that ATI's AF implementation retained lesser performance hit, even without off angles being present, which you, again, seem to have agree with above.

No, your theory is that the off angle problem stems from whatever process ATI uses when AF, is it not?

Yes. You're asking me to prove this, beyond observing that it happens and there is a performance hit reduction even when it does not? Well, the alternative is that the off angle problem is separate from the methodology, which is conceivable, but doesn't seem reasonable right now with no evidence anywhere of it ever being separated, and indication to the contrary of it being separable.

That is a consequence of the algorithms being used, so that in fact, they are essentially one and the same.

I don't get why you just repeat phrases like this like I haven't addressed them before. The result is not the same as the process.

Again: one output is one indication of the characteristics of the process and its input, it doesn't define the process by itself.

There is evidence, but there isn't the degree of certainty you've recently stipulated (at least, for what I've found). Why are your comments not subject to being evaluated for uncertainty? What do you think reasoning is, if not a way to work out which uncertainty is likely to be "certain"?

Because I never said my theory wasn't wrong. You seem to be saying that your theory must be right and my theory must be wrong(because yours is right)

I'm saying that specific indications that also happen to support my theory, as far I'm applying it, serve to work against your (sometime) theory that ATI only gains performance with the presence of off angles. This is not saying that your theory is wrong because mine is right, this is saying that something works to disprove that theory, and supports mine. There are other somethings that support both, or neither, as well.

This phrase, "after the fact"...how do you mean it? Such results support the theory. You can't just agree they support the theory in one place, and then go on about my not having provided results that support the theory in another.

General results support your theory, not specific ones. ie, you are looking at the lower performance hit of ATI's AF and saying that the off angles occur because of ATI's aggressive/efficient process.

You are abusing general and specific here. Consider the phrase "specific results generally support my theory".

Here is a contrasting "general" and "specific" case:

General: "The roof leaks when it rains".
Specific: "The roof is leaking right now, and it leaked saturday and sunday as well".

Now, you can say that the first statement is "general", and that the statement is "not specific". You can say that the first statement supports that you need to get your roof repaired. You can't pretend it never specifically rained for the "general statement", and propose that "specific cases" don't support that the same thing.

But we already agreed that most games don't use very many off angles, so using such data doesn't prove anything about off angles(to me).

I'm not sure why you continue to converse as if an indication can't serve to disprove your sometime theory about off angles. I.e., disprove that their presence is required to manifest a performance gain for AF.

In other words, the theory supports the evidence, more than the evidence supports the theory.

This looks like a bass ackwards statement, like "That was true because you said it, more than you said it because it was true", in response to someone saying something like "it is raining".

No, it is more like your theory does not fall apart when looking at general cases.

How did we arrive at these "general cases", then?

But when looking at specific cases, it might. But we don't have an specific cases to look at! ie, where off angles are isolated, that is a specific case.

This is what I mean about you continuing to demand that I provide "specifics" at one place, while agreeing with what you're demanding the "specifics" for, elsewhere.

OK, pardon me for quoting fully, but could you highlight the part of this text that says "the organization of the process and the process cannot be proven" instead of "the organization of the process and the process cannot proven by only measuring the result", except as you continue to equate "result" with "process", ignoring everything I've said on the subject?!

If it is not proven by looking at the result, then what is it proven by?

It isn't proven by looking at the one measure of the result in question. It is covered by looking at related information, and sitting down and thinking about it, and proposing a theory, and testing it. The entire logic thing.

Well, I'm only trying to get so far as to establish that they are not completely defined by counting texels, and that the count of texels is what is completed defined by counting texels.

Ok, that's fine. So "work" and the "process" are not completely defined by filtered texels. I thought we covered that.

OK, we did, for now.

I'm asking what "work" and the "process" are defined by. Do you not understand that?

Sure I understand that, which is why I gave you the answer before. Please note the word "complete" above, and seriously let it linger in your mind a bit...my making a point of it in various contexts, all having to do with why it is important, is related to this.

Since it seems necessary again:

You can measure a result of a process, and this defines that result completely (the texel count of a tunnel, for example). You can also measure different results of a process (the texel count of a "non off angle" surface). You can look at a different measure of some aspect of the process (for example, performance hit).

Now, you can pick one result you measured, and try to draw associations while ignoring other things that contradict them (for example, noting image quality degradation at off angles, ignoring whether performance hit depends on it, and concluding that performance hit is reduced solely by the presence of off angles, and therefore that measuring the work result for off angles represents how it gains performance). This is not complete.

You can look at other results, and if they include ones that invalidate the above associations, this consideration is complete as far as disproving those associations as they were stated. They were not, however, complete in defining the process...it is only complete in making progress towards doing so with the information considered so far.

Now, have I "defined" some characteristics of the process, and how they seem to apply to contradicting your initial statement? I think so. Have I defined the process? Not completely, but only so far. Is that a yes, or a no? As far as your requiring an answer justifying it as alternative to defining the process by looking at one specific result and failing to consider other aspects related to the process, it is a "yes".

You want a formula representation and analysis of how ATI implements their AF, then? What about your statements do you think requires me to specify such in order to be able to refute? What I've provided seems sufficient, and complaining that I haven't completed reverse engineering ATI's implementation doesn't do anything to address that.

No, I am asking for what variables(specifically, what these variables represent) would make up such formulas.

That would depend on how it calculates, but it would be related to representing the texture more correctly as perspective and orientation away from parallel to the screen plane change.

Since you say that filtered texels have are the result and not part of the process,

What if I looked at the results for "on" angles, and said that they defined the process? Does that mean that the Radeon AF always filters to provide more detail? I'm sure you'd say that outside of the "on" angle case behavior, this isn't the complete picture, and mention a tunnel test. I'm not sure why you have difficulty in grasping that outside of observing specific angle variance, the tunnel isn't complete either.

and that number of samples taken is not part of the process.

Hmm? Counting the samples taken as present in the output tells you the count of samples taken and then presented in the output, it doesn't define the process. It is part of evaluating the process, but not the end of evaluating the process. If you're measuring the samples evident in output, it doesn't even tell you directly the samples that were input, just the number of samples used as far as your methodolgy measures effectively.

What IS part of the process?

I'm not stating the mathermatical and organizational principles involved with ATI's AF implementation in particular, I'm describing characteristics about it and applying them to your statements that don't seem to fit.

An example. Is there something I should explain about what an example is, now?

A more specific example that tells me how these arbitrary numbers used in your previous example are related to AF?

They're not related to AF at all, they're related to the problems in your statements you continue to persist in. That is what I meant by asking if I had to explain what an example was.

If full degree AF was used at all angles then:
a) performance would globally decrease, whether or not former off angles are present.

If full degree AF was used at all angles by implementing things with the same process and implementation organization as the GeForce, yes.

b) performance would only decrease when off angles are present.

If full degree AF was formerly used at non-"off" angles by implementing things with the same process and implementation organization as the GeForce, and this was then changed to apply to off angles where it didn't before, yes.

So you are saying that it is possible to perform full degree AF at "off angles" where currently only 2x AF is performed, without losing performance in cases where such "off angles" don't exist.

This doesn't fit "a" at all, where performance "globally decreases", so you mean "b"? For "b", I'm saying that if the process calculated like the GF, but had a mechanism to specifically turn down the degree of aniso for off angles (i.e., it gained performance only by reducing image quality), that turning off the mechanism of detection would only decrease performance when off angles were present, because the calculation part of the process is the same, and the performance decrease would be the same when the detection mechanism wasn't in effect ("non off" angles).

You are also saying that it is possible to implement AF another way so that performance will globally decrease.

"a"? Yes, that is related to how some hardware does things more efficiently than others.

So....if it is possible to separate the off angle problem from the "normal" AF ATI is currently doing,

No, I stipulated that the answer to one of your questions was your initial proposition was the case. As we'd covered...?

why is it impossible that the off angle problem is unrelated to the way ATI is doing AF?

Ack!

I have said from the beginning that I thought the off angle problem was mututally exclusive from "normal operation" of AF, and you disagreed.

Double Ack! I'd have to repeat most of this post to dig out of this illogic, including covering things you'd already said you agreed with. If I can't even count on you to stick to your own statements, how can you hope to discuss anything in a non-circular fashion?

 

[StealthHawk]

I would be skeptical of any results which purport that treating all texture stages with trilinear has no more of an impact on performance than treating a single stage. There may be instances--in cpu limited games, for instance--in which this might appear to be true, no doubt. But it wouldn't be true because there is no performance hit from treating 4 stages as opposed to 1... I'd think the same way about someone telling me that 4x FSAA exacts no more of a performance penalty than 2x FSAA--that might be true in certain games--but not because 4x isn't demanding a lot more overhead...

Well, to be fair, in your example it is more like 1 stage trilinear 3 stages bilinear vs all 4 stages trilinear, rather than 1 stage vs 4 stages.

Has ATI ever commented on rTool/registry key that allows trilinear AF on all stages?

 

[StealthHawk]

If that is what I was addressing, well then, yes. And it was what I was addressing. A conversation, yes?

I have since stopped addressing it though :? If my position is fundamentally flawed, which according to you it is, why should we discuss it? My theory is reliant on less work being done on off angles, which you say is not true. Therefore, I am wrong. Nothing to discuss

Actually, you do have evidence to disprove the statement that the performance difference only occurred with the presence of off angles.

Are you reading my post in its entirety then responding? Because from the last few posts it looks like you are responding to it piecemeal, such as the above reponse. Later in my post, I already expanded and clarified on what I believe about ATI's AF performance.

I provide evidence. You respond by saying that I didn't, or that it isn't incontrovertible and final and therefore doesn't count. I point out directly to you why such a comment doesn't seem valid. You change your statement to say there is evidence after all, or that you never disputed that I had evidence, but then you go on to ask me to provide evidence again.

You have not provided evidence beyond conjecture and generalization. I have never asked for evidence that ATI's AF had a lower performance hit. I have always asked for evidence in relation to how ATI's AF does work, what it's process contains, or isolated conditions using off angles.

Hmm? The bias value itself, but not its effects, right? Unless the LOD bias is adjusted along with AF being used, to, for example, increase texture detail without AF processing as I think we discussed, or reduce aliasing while increasing AF processing.

So is LOD bias changing when AF is enabled or not

OK, but this statement, even with your stipulation, supports that "it does more work so it's better image quality" is an invalid precept, which was the point of my conversation.

I am not arguing with that. I am saying that it is not fair to apply it because of off angles. To which, I have stated I believe less work is being applied, and IQ is visibly decreases. Whether or not most games use off angles heavily or not, they are there to some degree. In some games more off angles are present than in others. In other words, I don't think the phrase is totally accurate(because some people will use it as an excuse to say that off angle problems don't exist and that ATI's AF is 100% superior to NVIDIA's in all ways. Yes, there are people like that).

Yes. You're asking me to prove this, beyond observing that it happens and there is a performance hit reduction even when it does not? Well, the alternative is that the off angle problem is separate from the methodology, which is conceivable, but doesn't seem reasonable right now with no evidence anywhere of it ever being separated, and indication to the contrary of it being separable.

Well, in R200 45 degree angles did not receive full treatment. But in R300 they do. R300's bilinear AF seems to have the same almost free performance hit as R200 I think. In fact, I have raised this point before.

That is a consequence of the algorithms being used, so that in fact, they are essentially one and the same.

I don't get why you just repeat phrases like this like I haven't addressed them before. The result is not the same as the process.

Again: one output is one indication of the characteristics of the process and its input, it doesn't define the process by itself.

You have said clearly that the off angle "issues" are a result of the process. You're saying that the AF alogrithms are not the process :?

General results support your theory, not specific ones. ie, you are looking at the lower performance hit of ATI's AF and saying that the off angles occur because of ATI's aggressive/efficient process.

You are abusing general and specific here. Consider the phrase "specific results generally support my theory".

Here is a contrasting "general" and "specific" case:

General: "The roof leaks when it rains".
Specific: "The roof is leaking right now, and it leaked saturday and sunday as well".

We are playing a semantics game here. "specific" results refers to specific benchmark results, it does not mean that said results have the characteristics to make them specific to the topic at hand. So the results are "specific," but they are as they are related to off angles and ATI's AF, they are general.

I'm not sure why you continue to converse as if an indication can't serve to disprove your sometime theory about off angles. I.e., disprove that their presence is required to manifest a performance gain for AF.

I thought I already said that ATI had a lower performance hit I don't know where the hell you got the idea that I said ATI's performance hit wasn't lower. As far as I remember, I have NEVER said either of the following:

1) ATI has the same performance hit as NVIDIA when no off angles are present
2) ATI does the same amount of work as NVIDIA when no off angles are present

It seems you think I've said, one or both of these though :? Please show me where I said that.

How did we arrive at these "general cases", then?

Because there is no data where off angles are isolated or even meeasurably present, and that is what is under dispute. Hence, what you are providing is "general" evidence about ATI's AF.

I don't see what is so hard about this to understand. Suppose you have an apple orchard where you grow red and green apples. Suppose you pick 15 apples, 5 green and 10 red. You tell me that the red apples weigh twice as much as the green apples, and you ask me how much each green apple weighs. I can't tell you because I don't know the total weight of the 15 apples. In this case, I know some specific information about red apples, but this information is totally unhelpful in determining anything wrt all of the apples together or the green apples.

Double Ack! I'd have to repeat most of this post to dig out of this illogic, including covering things you'd already said you agreed with. If I can't even count on you to stick to your own statements, how can you hope to discuss anything in a non-circular fashion?

As said somewhere above, it seems that you think I've said quite a few things that I never did, and I have done the same to you.

Ok, another agree or disagree time; let's say that ATI changes their AF implementation:

1) ATI cannot perform full degree AF at all angles without losing performance even when "off angles" are not present relative to their old(current) method.

2) ATI can perform full degree AF at all angles without losing performance even when "off angles" are not present relative to their old(current) method.

 

[demalion]

If that is what I was addressing, well then, yes. And it was what I was addressing. A conversation, yes?

I have since stopped addressing it though :? If my position is fundamentally flawed, which according to you it is, why should we discuss it?

Because you continue to propose things based on it not being flawed.

My theory is reliant on less work being done on off angles, which you say is not true.

No, your theory is dependent on off angles being present being the only way to do less work. Are we on the same page? What does "performance difference only occured with the presence of off angles" mean to you? It is consistent with some of the things you say, but not others. The point of my conversation was addressing the things it is consistent with, already highlighted as part of the discussion about your self-contradiction. Yet again I ask, how much clearer can I be?

Therefore, I am wrong. Nothing to discuss

Then why do you keep proposing disagreement to my "theory" based on the same problems?

Actually, you do have evidence to disprove the statement that the performance difference only occurred with the presence of off angles.

Are you reading my post in its entirety then responding? Because from the last few posts it looks like you are responding to it piecemeal, such as the above reponse. Later in my post, I already expanded and clarified on what I believe about ATI's AF performance.

What don't you understand about the phrase "contradiction", and the associated demonstrations of it in your text that I've provided? Do you think by simply failing to respond to my discussion of it, as you've done several times, makes it go away? Alternatively, perhaps this self-contradiction is related to why I was correcting the statement you were making here, and why your bringing up a statement you made elsewhere is nonsensical?



Self-contradiction and incorrect statements completely negates any ability for us to converse when you simply ignore them when I point them out to you, and then you continue conversation based on them.

For instance, the entire "General results support your theory, not specific ones" issue, where you just repeat that accepting the general case supporting something means that specific cases do not, and simply ignore that a "general case" does not exclude all specific cases because "general" and "specific" mean different things.

I provide evidence. You respond by saying that I didn't, or that it isn't incontrovertible and final and therefore doesn't count. I point out directly to you why such a comment doesn't seem valid. You change your statement to say there is evidence after all, or that you never disputed that I had evidence, but then you go on to ask me to provide evidence again.

You have not provided evidence beyond conjecture and generalization. I have never asked for evidence that ATI's AF had a lower performance hit.
...

Do you see the problem here? My actual quoting and discussion of your statements simply disappeared from the discussion.

Hmm? The bias value itself, but not its effects, right? Unless the LOD bias is adjusted along with AF being used, to, for example, increase texture detail without AF processing as I think we discussed, or reduce aliasing while increasing AF processing.

So is LOD bias changing when AF is enabled or not

The bias can change when AF is enabled. The bias can remain unchanged when AF is enabled. The bias can change, even without AF being turned on or off. Your question confuses causality. We've had this conversation.

OK, but this statement, even with your stipulation, supports that "it does more work so it's better image quality" is an invalid precept, which was the point of my conversation.

I am not arguing with that. I am saying that it is not fair to apply it because of off angles.

Are you introducing some sort of "political logic"? What do you mean by "not fair"?

To which, I have stated I believe less work is being applied, and IQ is visibly decreases.

Image quality is visibly decreased, and less work can be said to be applied, yes, as long as you aren't proposing a specific relationship between them in this statement.

Whether or not most games use off angles heavily or not, they are there to some degree. In some games more off angles are present than in others. In other words, I don't think the phrase is totally accurate(because some people will use it as an excuse to say that off angle problems don't exist and that ATI's AF is 100% superior to NVIDIA's in all ways. Yes, there are people like that).

If preventing people from making incorrect statements were enough justification to exclude observations, it would be just as much justification for saying the ATI's AF doesn't have an off angle problem to prevent people from saying that image quality always results from doing more work. I know you see the problem with this, but I don't know why you don't see fit to prevent yourself from making statements based on this premise. Another conversation we've had before.

OK, skipping to more direct questions, in hopes of progress:

That is a consequence of the algorithms being used, so that in fact, they are essentially one and the same

I don't get why you just repeat phrases like this like I haven't addressed them before. The result is not the same as the process.

Again: one output is one indication of the characteristics of the process and its input, it doesn't define the process by itself.

You have said clearly that the off angle "issues" are a result of the process. You're saying that the AF alogrithms are not the process :?

No, I said clearly that the result of the AF alogrithm is not the process.

As far as I remember, I have NEVER said either of the following:

1) ATI has the same performance hit as NVIDIA when no off angles are present
2) ATI does the same amount of work as NVIDIA when no off angles are present

It seems you think I've said, one or both of these though :? Please show me where I said that.

Well, one place I've quoted already, with highlights, and you then just decided to skip over:

We need to be clear on this: does ATI's performance hit being reduced depend on the off angles being present, or not?

Yes, I think so.

Also, in your initial proposal in disagreement to Quitch, and throughout our conversation. How can you say I've debunked your theory in one place, and then say that you never had the theory in the first place in another? Are you just ignoring the majority of my discussion which establishes this, and then wondering why I say there is a problem with the way you're responding to my explanations and questions?

How did we arrive at these "general cases", then?

Because there is no data where off angles are isolated or even meeasurably present, and that is what is under dispute.

What data about off angles do I need to present to prove that ATI's being more efficient even without them, when you've agreed that it is already? This is one of your contradictions. Please take another look at our conversation, with more care than the first time, as I've requested several times.

Hence, what you are providing is "general" evidence about ATI's AF.

I don't see what is so hard about this to understand. Suppose you have an apple orchard where you grow red and green apples. Suppose you pick 15 apples, 5 green and 10 red. You tell me that the red apples weigh twice as much as the green apples, and you ask me how much each green apple weighs. I can't tell you because I don't know the total weight of the 15 apples.

Well, it is more like asking you how many green apples would be required to approach the weight of the 10 red apples. You could answer 20, based on the specific indications you have. We haven't established for sure that the next 15 green apples will weigh the same as the first 5, but that doesn't mean that the specific weight of the 5 green apples don't support this because we represented the information as "general".

In this case, I know some specific information about red apples, but this information is totally unhelpful in determining anything wrt all of the apples together or the green apples.

And this evaluation seems totally incorrect. What it is not is the end of determing the complete information about all of the apples together or of the green apples.

If someone pointed out that their relative weights corresponded to their relationship in size, you couldn't ignore it to propose that red things are heavier than green things, without also necessarily proposing other things that are not supported while simultaneously ignoring that the size determing the weight is a more reasonable statement with based on "general" information.

...
Ok, another agree or disagree time; let's say that ATI changes their AF implementation:

What is the point when you ignore my answers? Ok, as more one way courtesy...

1) ATI cannot perform full degree AF at all angles without losing performance even when "off angles" are not present relative to their old(current) method.

Well, they seem to have provided indication against this for 45 degree separated angles already. I can't define an answer that precludes them from doing so further in the future, or evaluating how it will compare to the current implementation. Disagree.

If they switched to doing it the same way as the GeForce has, which is one possibility, I'd agree.

2) ATI can perform full degree AF at all angles without losing performance even when "off angles" are not present relative to their old(current) method.

Hmm? If they changed their implementation, it is possible. It is also possible that they couldn't perform full degree AF at all angles without losing performance compared to their prior method. This would depend on what the new process, that achieved full degree aniso at all angles, was.

 

[Bjorn]

R300's bilinear AF seems to have the same almost free performance hit as R200 I think. In fact, I have raised this point before.

Well, there's actually some indications that it might have a larger performance hit. 2X -> 16X AF = 14% perf hit for a 9600 Pro vs 6% between 0X and 16X AF for the R9000 Pro. But this is not with the same demo (both are UT2003 though) and then there's of course other things that play a role in this as we have seen (diff between 9600/9800).

 

[StealthHawk]

So is LOD bias changing when AF is enabled or not

The bias can change when AF is enabled. The bias can remain unchanged when AF is enabled. The bias can change, even without AF being turned on or off. Your question confuses causality. We've had this conversation.

This is not an answer. Is LOD bias changing when AF is enabled or isn't it? Here, just like other places, you seem to give me the runaround, rather than providing a straight answer.

Earlier you seemed to be saying that LOD was changing, or maybe you were just making a proposition

Why wouldn't the texture aliasing "people claim" be related to a higher level of detail for textures (the default for Direct3D is "highest" last I checked)?

[quute]

That is a consequence of the algorithms being used, so that in fact, they are essentially one and the same

I don't get why you just repeat phrases like this like I haven't addressed them before. The result is not the same as the process.

Again: one output is one indication of the characteristics of the process and its input, it doesn't define the process by itself.

You have said clearly that the off angle "issues" are a result of the process. You're saying that the AF alogrithms are not the process :?

No, I said clearly that the result of the AF alogrithm is not the process. [/quote]

Where did I say in the first quote that the result is the process I don't see that at all.

As far as I remember, I have NEVER said either of the following:

1) ATI has the same performance hit as NVIDIA when no off angles are present
2) ATI does the same amount of work as NVIDIA when no off angles are present

It seems you think I've said, one or both of these though :? Please show me where I said that.

Well, one place I've quoted already, with highlights, and you then just decided to skip over:

We need to be clear on this: does ATI's performance hit being reduced depend on the off angles being present, or not?

Yes, I think so.

Way to selectively quote, here is what I said in its entirety.

Yes, I think so. While I have admitted that ATI's algorithm is more efficient/aggressive than NVIDIA's, when there are off angles present that workload is going to decrease even more in comparison to the work being done by NVIDIA.

I have already clarified that as well. So again, where have I said that ATI's AF does not have a lower performance hit than NVIDIAs Because I'm pretty sure I've never said that. I don't understand why you keep insisting that is what I've been saying.

Also, in your initial proposal in disagreement to Quitch, and throughout our conversation. How can you say I've debunked your theory in one place, and then say that you never had the theory in the first place in another? Are you just ignoring the majority of my discussion which establishes this, and then wondering why I say there is a problem with the way you're responding to my explanations and questions?

Apparently yes I am ignoring it, since you like to play semantics and take quotes out of context. Either that, or my articulation is egregiously bad, to the point that every other word I type is being misconstrued.


I just don't understand how you expect me to seriously consider your theory when you refuse to give straight answers about it. Obviously you do not know exactly how ATI's AF works. But you must have some belief as to how you think it works.

Furthermore, when you just said that 45 degree angles are special cases. How then is this compatible with your theory that the same amount of work is being done at all angles. Doesn't a "special case" in and of itself imply that something different is being done when this type of case comes up...

I don't think this conversation is going anywhere, this is no longer enjoyable for me, and I am waiving my obligation to reply to any future posts on this subject unless something interesting comes up or I feel progress will be made. I thought at some points we were making progress but as you've said, we have come full circle already. You keep referring me to past posts and now I'm doing the same :shudder:

 

[WaltC]

Well, to be fair, in your example it is more like 1 stage trilinear 3 stages bilinear vs all 4 stages trilinear, rather than 1 stage vs 4 stages.

Has ATI ever commented on rTool/registry key that allows trilinear AF on all stages?

I'm thinking though that bilinear is the chip default for stage treatment (which might lead some to propose it's "free bilinear.") I haven't seen a point-sampling option in a long time... Not sure about that, however. As to rTool, I don't know that they generally comment on the efficacy (or lack thereof) of tweakers...

 

[demalion]

So is LOD bias changing when AF is enabled or not

The bias can change when AF is enabled. The bias can remain unchanged when AF is enabled. The bias can change, even without AF being turned on or off. Your question confuses causality. We've had this conversation.

This is not an answer. Is LOD bias changing when AF is enabled or isn't it?

"Not necessarily" is my answer, even if you don't like it, or whatever your objection is. I definitively answered the question of whether "LOD bias setting HAS to change with AF", which is all that was required to address your contention that "LOD bias has to be part of AF and texture aliasing being present means it was an issue with ATI's AF"!
What is this game with ignoring that my answer already got my point across, to make up a new question defining some new tangent? The only point I see is to waste my time and ignore that I've already established the problem with your statement.

"Will it be raining outside of my house at 8:00 AM tomorrow?" If you care about your answer being as true as you know how, perhaps you see the issue with your question if you try to answer this? In any case...according to you, what is the purpose of this new question?

For instance: I have a recollection sometime in the GeForce driver history, LOD bias did change with the AF level selected. I also have a recollection that some time in ATI driver history, LOD bias did change with something known as an "AF level" in the registry (the so called "32, 64, and 128x anisotropic filtering" options). Does this mean it changes now in the same way? Does this mean it does or doesn't change to reduce texture aliasing as part of an implementation? No, but these answer have nothing to do with pointing out that LOD bias being too aggressive might be the cause of aliasing, or change that you can turn down the LOD bias separately from the AF setting.

Here, just like other places, you seem to give me the runaround, rather than providing a straight answer.

Runaround!? SH, I don't know what your problem is. Are you really paying so little attention to what I'm saying? Maybe you honestly don't realize your initial comment was already addressed, and this was my entire point in...replying to your initial comment?!

Earlier you seemed to be saying that LOD was changing, or maybe you were just making a proposition

Why wouldn't the texture aliasing "people claim" be related to a higher level of detail for textures (the default for Direct3D is "highest" last I checked)?

Eh? Where do you get "LOD was changing" out of that?

"the default for Direct3D is "highest" last I checked". This means that the default for ATI is the "Highest" mip map level selection (and Highest texture quality, though I don't currently think this boosts LOD bias). The driver's behavior with this setting, as well as the game's LOD settings (for instance, UT2k3 commonly uses negative LOD settings), could be responsible for texture aliasing. Therefore, the result being texture aliasing, doesn't tell you what the cause was, and a possible cause is something besides the AF implementation.

OK, why did I have to explain that?

That is a consequence of the algorithms being used, so that in fact, they are essentially one and the same

I don't get why you just repeat phrases like this like I haven't addressed them before. The result is not the same as the process.

Again: one output is one indication of the characteristics of the process and its input, it doesn't define the process by itself.

You have said clearly that the off angle "issues" are a result of the process. You're saying that the AF alogrithms are not the process :?

No, I said clearly that the result of the AF alogrithm is not the process.

Where did I say in the first quote that the result is the process I don't see that at all.

Now we're playing games with "algorithm" and "That is a consequence".

First, both of your sentences: "No, your theory is that the off angle problem stems from whatever process ATI uses when AF, is it not? That is a consequence of the algorithms being used, so that in fact, they are essentially one and the same."

Hey, your "That" is referring to the off angle problem, which, as I've discussed, is a result. It seems pretty obvious that the AF algorithim is the process. It also seems pretty obvious what "one in the same" means. How you fail to see how this associates with "Result is the process", despite my explaining all of these items several times, is not obvious, and not even interesting. It is (IMO) rude, lazy, and/or associated with some agenda directed personally at me, or your perception of me.

Maybe you think I'm an "ATI zealot" or something, and got stuck on the idea that the issue was debunking things "ATI zealots" might do, as you alluded to earlier in your discussion on "fairness". I don't know, and I don't see the point in wasting more time in trying to find out if it could be something different when you so consistently seem to only be interested in focusing on issues that only make sense in some sort of political regard, instead of one that is logical or factual.

To cover remaining highlights:

We need to be clear on this: does ATI's performance hit being reduced depend on the off angles being present, or not?

Yes, I think so.

Way to selectively quote,...

Yes, because I selected the contradiction.

...here is what I said in its entirety.

Yes, I think so. While I have admitted that ATI's algorithm is more efficient/aggressive than NVIDIA's, when there are off angles present that workload is going to decrease even more in comparison to the work being done by NVIDIA.

Self-contradiction.

contradiction.

What don't you understand about the phrase "contradiction", and the associated demonstrations of it in your text that I've provided? Do you think by simply failing to respond to my discussion of it, as you've done several times, makes it go away? Alternatively, perhaps this self-contradiction is related to why I was correcting the statement you were making here, and why your bringing up a statement you made elsewhere is nonsensical?

If you were consistent, the answer to my question, "does ATI's performance hit being reduced depend on the off angles being present, or not?", would have been "No. While...", instead of "Yes, I think so. While...", and the conversation could have progressed without there being confusion in how this answer of yours related the rest of your questions in that post.
Instead, we have one part of your conversation saying "No, ATi's doesn't depend on off angles being present for its performance hit being reduced", and another part saying "Yes, ATi depends on the off angles being present for its performance hit being reduced".

This is why I chose the word "self-contradiction", and similar issues are why I discussed it with you in several posts running. This is just one clear example...for others, pease actually read what I've said when I characterize something you've said as contradictory.
Perhaps the link to an explanation of the meaning of the words will help.
I have no interest in doing the work of finding every instance, and have you respond as you have consistently done to my efforts to answer your questions and provide explanations in this thread, so I also look forward to whatever alternative to that occurs.
If for some reason this reply finally sparks some recognition of what I've been saying, I can provide further details and references to things I've stated already in PMs.