New 3DMark03 Patch 330

[Pete]

Speaking of 3DM01, I can't help but link to this picture:

Cheating Drivers, Hidden Dragon

Simply the best 3D spoof I've seen.

 

[demalion]

Walt-

What is it with you and these "post filter" conspiracy theories? There is really nothing suspicious or untoward about using a post filter: a post filter is capable of doing anything that a filter on the GPU can do; the only difference is its location, both physically on the chip and schematically in the rendering process. Under the simplest method of using a post filter to filter MSAA subsamples, you just have a tradeoff of less bus utilization (because you save the work of transfering the frontbuffer to the GPU for filtering and transfering back the filtered data to serve as the backbuffer) in exchange for a larger memory footprint (because for n-way MSAA your unfiltered backbuffer still has n sub-samples for every pixel).

"Conspiracy theory?" *chuckle* I don't recall objecting to nVidia using the post filter in any way they choose

I do happen to recall that, and the difficulty in spotting new support in your replies might have Dave H recalling the same thing:

WaltC[/url]]nVidia doesn't want it known that their "2xFSAA and Quincunx" numbers are frauds that don't involve FSAA at all--but involve only post filter stuff--which is why the screen shot software doesn't pick it up.
...

I do not, however, know why you insist in shrouding your discussions with derisive commentary and *chuckle*-laden condescension. Makes it harder to discern the point.

--what I object to is them refusing to describe it in any terms whatsoever, claiming it is a "trade secret" so incredibly valuable that even mentioning it in a generally descriptive way would endanger the "trade secret." That's a perfectly valid point.

That is a valid point of objection, I agree, it is just unfortunate that there is so much unproductive commentary to go with it. Let me try to help circumvent that by proposing an actual theory on your behalf: The RAMDAC's color precision for downsample calculation is less than the downsample precision that is used elsewhere.

It isn't the only theory (they could just be lowering texture detail to boost benchmark scores, which has the benefit of not requiring a further major screw-up on nVidia's part), but it is one theory that fits some observations and satisfies your evident predilection for condemning postfiltering.

I think it would be productive if your text contained an addressable theory like this (like Dave H's text refers to an addressable theory) among the exhibited dedication to chuckling and condescending accusations, so feel free to use it or a modification in further discussion please.

Look, you claim to know what they are doing and how--will you link me to a source from nVidia corroborating your comments?

Why should he be required to do more when you don't even do as much as he's already done? He is proposing a theory that holds together, so far, and providing logical support. It also doesn't happen to necessitate precluding indications reported by people. It isn't the only theory that could fit that criteria, but you aren't proposing an alternate theory in opposition, just attacking that theory, with the only support being to send a barrage of *chuckle*s and winks his way.

It would be much appreciated. *chuckle* The only "conspiracy" I see here is people talking about it in the absence of any released information on the subject from nVidia...For a subject nVidia has quite literally clammed up on--you seem to know a lot about it...

Good grief, Walt, please try to stick to disagreement with a point...you're just spending your time making insulting statements to belabor that nVidia hasn't verified exactly what they are doing, while ignoring that your attacks are not further validated by that. This circumvents that his proposition is not disqualified by observations provided as of yet...but if you really think that it is, please just say so more directly.

BTW, the "trade secret" thing is not my invention--that's a quote of the words nVidia used when it declined to talk about what it was doing any further. Your questions are better directed toward nVidia than me...most assuredly... I'm only guilty of wanting to hear it from the horse's mouth...

By your tactic of "argumentation", this would be the time to drop in a comment about another part of a horse's anatomy to describe your behavior. Wouldn't that be productive? I really find many of your posts incredibly grating, as I've said before.

I know you can structure a solid argument, but it seems you enjoy insulting people way too much to bother very often, without being pressed on the matter.

Although, the last time we talked this over Demalion did a much more detailed analysis which IIRC seemed to show that the post filter version actually consumed more bandwidth in the >2x MSAA case, although I forget why. (I had to have it explained to me slowly back then, too...) But in any case, that's the basic idea; and it should be obvious that the same filtering function can be used in a post filter as in the normal way of doing things.

How can *anything* on the subject be "obvious" when nVidia refuses to talk about it...?

Well, what he said was that it is "obvious that the same filtering function can be used in a post filter as in the normal way of doing things." If you follow the above link and continue reading, you'll even note prior explanations as to how this recognizes the possibility that it might not do this equivalently, even though it certainly could. You are attacking the idea of his certainty, while not recognizing that any certainty on your part is less supported, and you are doing that by making the mistake(?) of interpreting his comment as a certainty about what nVidia is doing rather than being a certainty about what they could do. I think that would be your fault, because the latter is what he actually stated...you haven't answered that in your reply, or stated why it is invalid for him to propose it as an explanation for what they are doing in actuality.

I mean we can do conjecture and spin hypotheses all day long--doesn't mean they're correct or even pertinent.

What's your hypothesis? Perhaps my annoyance caused me to miss it and its support, but it just seemed to me your argument consisted only of (falsely) saying that Dave's hypothesis and support for it was only based on a certainly of what nVidia is doing. That argument is not a hypothesis, that's just recognizing that an alternate hypothesis might not be a fact.

I'm just saying it would be nice for nVidia to declassify this "trade secret" (nVidia's words, not mine) and talk about it, explain its benefits, drawbacks if any, etc. That seems an entirely reasonable request.
...

Yes it does, but all the rest of your commentary that goes along with that request doesn't seem very reasonable at all, IMO.
For example, the rest of your post seems to completely ignore the support for the hypothesis already presented, because most, if not all, of the questions and statements you make in it are addressed by that hypothesis. What would be more productive, IMO, is responding to those parts of the hypothesis directly, instead of making statements and questions and ignoring the proposed answers to them. In your "levity", you seem to me to have failed to give thos parts any recognition in your reply, and that just doesn't seem useful.

 

[kyleb]

Speaking of 3DM01, I can't help but link to this picture:

Cheating Drivers, Hidden Dragon

Simply the best 3D spoof I've seen.

bwahahahahaha!

 

[mczak]

                Build  GT1   GT2  GT3
Matrox Parhelia 3.2.0  78.7  3.5  4.5
                3.3.0  78.9  3.5  4.5

Xabre 400       3.2.0  43.7  2.9  2.4
                3.3.0  43.5  2.9  2.4

so matrox and sis don't cheat in 3dmark03 (or at least not cheat in a way which is no longer possible, though given the scores it's probably safe to say they don't cheat at all). Can't say I'm surprised that Matrox did not cheat. SiS OTOH probably would if they would have a chance to beat a competitor (radeon 9000 or GFFX 5200) but their score is so terribly low that probably no amount of tweaking would help... Though Matrox too wouldn't gain much from cheating, even if they could get the scores up to radeon 9000 level, as the parhelia isn't a direct competitor for those "low-end" 3d cards.

 

[Dave H]

Walt-

So if I understand your reply correctly, the reason you keep harping on the (potential) post filter aspect of NV35's 4xMSAA implementation is that Nvidia said the details of NV30's post filter were a "trade secret". You think anytime they don't disclose the details, something suspicious is going on.

Well, I've got news for you: nearly everything in the GPU industry is considered a trade secret. Cache sizes, shader pipeline organization (ok, ATI disclosed the number of functional units--which Nvidia of course did not; but neither discloses compiled shader op format, latency/throughput for various ops, number of registers, etc.), sampling algorithms for various anisotropic filtering settings, mipmap selection algorithms for AF settings...hell, I don't even think they disclose the mipmap selection algorithms they use with AF turned off (yes, they're surely very close to the algorithms given in the OpenGL/DirectX specs...but R3xx and NV3x's algorithms differ slightly, so at least one much differ from the algorithm in the spec). Nvidia's not disclosing the way they filter MSAA sub-samples? Well, AFAIK ATI hasn't disclosed the way they filter their 6xMSAA sample pattern (not that this is a terribly difficult problem, but, unlike with their 2x and 4x patterns, weighting each sample equally would not produce optimal results in this case as I understand it).

And, contrary to the assertion in your original post, AFAICT 3dfx never disclosed any details about the 16->22bit upsampling post filter in the V3. You may have thought they did, because Kristof did such a damn convincing job exploring how such a feat could feasibly be done. But 3dfx never confirmed that he was right, or even on the right track. If you read those articles, you'll see that the closest they came was an email from Gary Tarolli:

Gary Tarolli[/url]]Congrats! You've written the best article yet. Your article was speculative, but your methods were very good, and your analysis was very thorough. That is, you were objective, not religious, like others.

No disclosure whatsoever of what was actually going on, just a hat tip to Kristof for having come up with a plausible guess. (And, BTW, they had a similar upsampling post filter with Voodoo2, but it caused horizontal artifacts, as detailed in Kristof's articles.)

Do I like the fact that everything about graphics hardware is considered a trade secret? No; I really dislike it. It's completely different from the situation in the CPU industry, where many details at a comparable level are disclosed. And it's pretty stupid, since many of the questions I listed have been answered (to greater or lesser degrees) by informed speculation, like Kristof's on the 22bit thing, and clever analysis programs, like the programs written by B3D forum members to determine FSAA sample placement and angle-dependent AF algorithms (or, more precisely, the mipmap selection algorithms that accompany them). Certainly Nvidia isn't keeping anything secret from anyone at ATI, nor vice-versa.

But they both play the secrets game anyways. Just because Nvidia is playing it with regard to their post filter doesn't make it any more or less suspicious than anything else in their products.

That's funny--so SS is not only a lot slower than MS, it produces a much inferior level of image quality, too? That's what you're saying...? Because that's what you get when you compare the two. That's exactly why I think the example is so telling (if you have the eyes to see it...)
Not only is 8xFSAA on the 5900U *much slower* than R9800P's 6x FSAA, it's also visibly inferior in the IQ it generates. Again, that's what makes the comparison noteworthy.

Sigh. I find it difficult to believe that you could hang out in these forums as much as you do and not understand the issues involved. But, to humor you...

No, supersampling does not result in inferior image quality comapred to multisampling. The reason Nvidia's 8xS has worse IQ than ATI's 6xMSAA is because ATI's implementation has better sample placement and gamma-corrects the subsample colors. It has nothing to do with the fact that one is multisampling while the other mixes multisampling and supersampling.

Moreover, supersampling has certain IQ advantages over multisampling + AF. For one thing, supersampling can antialias pseudo-edges caused by the use of alpha textures to mimic geometry detail in many older games (i.e. for leaves, ladders, etc.); multisampling + AF cannot. (Although many newer games use an AF-friendly version of the technique, albeit one which requries back-to-front rendering order.)

This makes NV3x's hybrid SS/MS modes very attractive for older games that use alpha textures: it provides much better IQ than R3xx's MS-only AA, and because the games are so old the performance difference is totally irrelevant. Games like this (Counterstrike is the most commonly cited example) are the reason for 8xS mode; comparing it to R3xx's 6xMSAA in, for example, UT2k3 is utterly beside the point.

To make a forward-looking version of the same point; AF (as currently implemented) only samples color textures, not the output of pixel shaders; thus using MS + AF will start to leave you with aliasing inside polys if high-contrast pixel shaders are used. Supersampling, of course, will antialias such situations (although given how bad NV3x's shader performance already is, one shudders to consider adding supersampling to the burden...).

 

[WaltC]

Walt-

So if I understand your reply correctly, the reason you keep harping on the (potential) post filter aspect of NV35's 4xMSAA implementation is that Nvidia said the details of NV30's post filter were a "trade secret". You think anytime they don't disclose the details, something suspicious is going on.

Dave, you keep misconstruing me on this issue, and this will be the third time you have heard me say: I don't care what nVidia does with the post filter relative to FSAA. I would simply like them to *generally* expound on it. I'm not asking for circuit schematics.... I hope we are clear on this, at long last.

Logically, it's very simple: if people have questions as to your methods to obtain stated results, it's good policy to answer such questions, and they can be answered in such a way that no valuable trade secrets are divulged. Very simple.

Well, I've got news for you: nearly everything in the GPU industry is considered a trade secret. Cache sizes, shader pipeline organization (ok, ATI disclosed the number of functional units--which Nvidia of course did not; but neither discloses compiled shader op format, latency/throughput for various ops, number of registers, etc.), sampling algorithms for various anisotropic filtering settings, mipmap selection algorithms for AF settings...hell, I don't even think they disclose the mipmap selection algorithms they use with AF turned off (yes, they're surely very close to the algorithms given in the OpenGL/DirectX specs...but R3xx and NV3x's algorithms differ slightly, so at least one much differ from the algorithm in the spec). Nvidia's not disclosing the way they filter MSAA sub-samples? Well, AFAIK ATI hasn't disclosed the way they filter their 6xMSAA sample pattern (not that this is a terribly difficult problem, but, unlike with their 2x and 4x patterns, weighting each sample equally would not produce optimal results in this case as I understand it).

Again--I would like generalties Dave--nVidia talking about what it's using the post filter for relative to its FSAA modes is no more a trade secret than nVidia revealing that in some FSAA modes it uses a combination of MS and SS, etc. Not a big deal at all. What I'm saying is not hard to follow.

nVidia's real big on PR, though--speaking of cache sizes *chuckle*--regarding nv30 I read a piece written by someone at nVidia which claimed that the 5800U's nv30 "internal cache bandwidth" was sufficient to increase the 5800U's "physical bandwidth" by 4gbs/sec (which just coincidentally then put the 5800U in the same physical bandwidth area as the R9700P.) And of course, lets not forget the initial PR reports about 60gbs/sec for nv30 that, as it turned out, were figures relating not to physical bandwidth at all but rather to assumed and estimated bandwidth as a result of compression occuring in some cases in FSAA.

Look, you might be satisfied with nVidia's PR--but I'm not. I would like to know.

And, contrary to the assertion in your original post, AFAICT 3dfx never disclosed any details about the 16->22bit upsampling post filter in the V3.

That is totally erroneous: Kristoff's article--or the one I remember--dealt with the difference in screen shots some web sites were taking and the screens shots which accurately portrayed V3 on-screen image quality. The post filter was examined from there and it's impact on screen-grabbing software explained. But look--Kristoff did a fine job with those topics--lets not blur things by dragging him into this discussion.

You keep talking about "detail"--what I want out of nVidia is generalitites--such as the ones 3dfx was very up-front about with the V3--up-front to the degree that they pushed their 16/22-bit mode as a *feature* (don't tell me you don't remember.)

What 3dfx revealed about their use of the post filter with the V3 was that (parphrased) "We are using the post filter, along with certain algorithms of our own, to blend our 16-bit output so that it gets very close to a 22-bit output on screen, but at 16-bit levels of performance." It was a feature they *pushed*--certainly not one they refused to describe generally as a "trade secret." If you think they did not push their use of the post filter and the resulting 16/22-bit mode you would be very much mistaken. In fact, the directly compared it to the image quality you could obtain on current nVidia products--TNT/2--which were much slower running in 24-bit mode, yet produced (at best) only slightly better results visually than the V3's 16/22-bit mode. I had a TNT and a V3 at the time and remember it so well because I was involved in the discussions about the relative merits of each--and at the time 3dfx's claims for the post filter advantages over TNT/2-24-bit display were justified (mainly because Glide was still big then and most game engines be they Glide or D3d used no better than 16-bit integer precision.)

Knowing that this is the purpose for which 3dfx employed the post filter, and knowing that 3dfx and all its IP are now the property of nVidia, and knowing that *unlike any of its previous products* nv3x board designs use the post filter for *some aspects* of their FSAA output--it is *natural and logical* to want an explanation from them *in general terms* as to why they are using the post filter for *some* of their FSAA modes.

Until I hear otherwise from nVidia, I will assume it is for the same reason 3dfx employed it--to enhance performance of, in this case, the nVidia FSAA modes which employ the post filter while doing FSAA. I think this is a very legitimate area of inquiry.

No, supersampling does not result in inferior image quality comapred to multisampling. The reason Nvidia's 8xS has worse IQ than ATI's 6xMSAA is because ATI's implementation has better sample placement and gamma-corrects the subsample colors. It has nothing to do with the fact that one is multisampling while the other mixes multisampling and supersampling.

*chuckle* I was addressing the image quality because in your previous response you totally avoided the subject, as if to say the only important variable was SS vs MS. And as you say--it's not... The whole point to an FSAA mode however it is exectued is IQ--there is no other point beside that. Questions of MS vs SS, RGSS vs. ordered, are all questions which take a back seat to the ultimate IQ of the FSAA mode as it is executed in various hardware. I see nothing in your comments which changes that. It certainly does not represent a "reason" for the maximum quality settings as established by each manufacturer to differ so much in IQ *and* performance. I will stipulate however that nVidia has never been very good with FSAA of any type, IMO.

Edit:

Recall the [H] review of the nv30 (5800U)...It was [H] who first noticed that a comparison of the screen shots they took at 2x FSAA and 0x FSAA were visually identical. But, they also noticed that what they observed on the screen was visibly different from the screenshots they took at 2x FSAA (recall that nVidia was pushing its 2x FSAA like no tomorrow at nv30's debut--they were actually calling people's attention to their 2x FSAA mode performance.) With the proper screen-grabbing software [H] was finally able to grab frames which mirrored what it saw on the screen at 2x FSAA, and you could at last see differences between 2x FSAA and 0xFSAA.

Now, what this strongly implies is that the nv30 when set to 2x FSAA was in fact doing 0x FSAA, but was employing the post-filter in a blending operation which nVidia felt simulated 2x FSAA on screen. What happened at first for [H] was that since there was no FSAA actually taking place the initial screen shots they took at 0x and 2x FSAA were identical, because the only thing happening when the reviewer selected 2x FSAA was *post filter blending.* They selected 2x FSAA in those initial nv30 drivers but what they actually got was 0x FSAA & post-filter blending (to simulate 2x FSAA.) That seems pretty obvious and is proven by the fact that the initial screen shots which did not grab the post-filter effects were identical, or so [H] said at the time.

This is kind of weird to think about, but the same thing happened to 3dfx when it employed the post filter in the V3. The company was unaware until the product had hit the review circuit that specific frame-grabbing software was needed to capture the post filter blending, and more than one review of the product was published with screen shots that were far below the actual image-quality of the product on screen. This was fixed quickly of course, but not before some damage was done. Now here's nVidia overlooking exactly the same thing--and thus we know why the standard screen grabbing software completely missed the post-filter effects that seemed to, according to the data [H] released, completely comprise nVidia's "2x FSAA" mode (which seemed to actually be 0x FSAA + post-filter blending.) Just as with 3dfx & the V3, invoking the post filter does cause some performance hit--just not nearly as much of a hit as a true 22-bit output would have caused, or in nVidia's case a true 2x FSAA mode.

Doubtless this may have changed with latter Dets and the nv35--but to what extent is uncertain (at least to me.) If you think about that you might see why I'm interested in this line of inquiry.

 

[PVR_Extremist]

It hasn't flown over MY head. I understand it.

If I was in your position I would just leave [H]. You don't belong there, you belong to a site like Beyond3D.

My sentiments exactly, and while your at it, moon Kyle and let a rip roarer out in his direction from all of us a B3D

 

[WaltC]

....
Yes it does, but all the rest of your commentary that goes along with that request doesn't seem very reasonable at all, IMO.
For example, the rest of your post seems to completely ignore the support for the hypothesis already presented, because most, if not all, of the questions and statements you make in it are addressed by that hypothesis. What would be more productive, IMO, is responding to those parts of the hypothesis directly, instead of making statements and questions and ignoring the proposed answers to them. In your "levity", you seem to me to have failed to give thos parts any recognition in your reply, and that just doesn't seem useful.

D, sorry about missing your post here initially, but I just saw it, so I didn't want you to consider yourself ignored...

First all of, criticizing someone's inflection, body language, or in this case prose or style of writing, is not a valid criticism of their statements. You can't refute something someone says merely by objecting to them putting a or or or a *chuckle* into their writing. You can try to refute it on that basis of course, but you'll fail. Right? I think you know this.

It's not exactly "levity" that I try to convey as much as it is "lightheartedness"--the idea that while I'm "serious" about what I'm saying I do not however view it as life-threatening.

I'm sorry you feel as if I've ignored your points--but you seem to be ignoring mine... It's great that you'll theorize about the matter of post-filter blending relative to nv3x FSAA, but that's not what I'm after. I simply want nVidia to talk about it as opposed to ficticiously maintaining (in my view) that talking about it in general is akin to furnishing the public with circuit diagrams of their processors. I mean, they talk about *so much else* in general that I hardly see any difference here.

It's fine if a hypothesis addresses what I consider to be the issues relevant to me, but I don't want a hypothesis, I want nVidia to talk about the generalities of its use of post-filter blending relative to FSAA in its current products. I'm sure you can see the difference. We can promulgate hypotheses from here to Andromeda, but that doesn't change the fact that nVidia's not talking about it, does it? And so none of the hypotheses can be validated or discredited.

There's a simple thing anyone who has an nv58/5900 with the latest Dets can do, though, and that's test out the current FSAA modes with the same kind of screen-grabbing software that [H] used in its initial 5800U review--the software that did not capture the post-filter blending. I would do it myself only I don't have the hardware at my disposal.

With this screen grabbing software you could contrast shots taken at 0x FSAA with shots taken at QC, 2x and 4x FSAA and inspect them for differences, and then repeat using the screengrab software that captures whatever blending operations are being done in the post filter--then you could determine to what extent if any nVidia is using post filter blending in certain FSAA modes to simulate all, or part of, its FSAA effects. You might also use this method to determine if post filter blending is being used at all in any of the current products and drivers, and in which FSAA modes the post filter is employed (if it is being used in any of them.)

That's the kind of data I'm looking for. My original criticisms of this were ignited by [H]'s original 5800U review a couple of months ago, in which [H] stated it could detect no visual differences between its screen grabs at 0x FSAA and its screen grabs at 2x FSAA, although they could see a difference on screen. This smacks of post filter blending, and the fact that modified screen-grab software was required before the on-screen 2X FSAA IQ could be captured simply underscores the likelihood that this is what is going on (or at least *was* going on at the time.)

Is it still going on? If so, to what degree and in which FSAA modes?

If nVidia would simply *tell us*, much in the same way it tells us which of its FSAA modes are MS and which are MS & SS, and which used ordered grids as opposed to rotated grids, etc., we'd all *know* and there'd be no need for "suspicions" or "hypotheses," would there?

Edit: typos

 

[Evildeus]

Well i think otherwise. It's good to have someone with knowledge at [H] who is able to "educate" the mass.

It hasn't flown over MY head. I understand it.

If I was in your position I would just leave [H]. You don't belong there, you belong to a site like Beyond3D.

My sentiments exactly, and while your at it, moon Kyle and let a rip roarer out in his direction from all of us a B3D

 

[demalion]

....
Yes it does, but all the rest of your commentary that goes along with that request doesn't seem very reasonable at all, IMO.
For example, the rest of your post seems to completely ignore the support for the hypothesis already presented, because most, if not all, of the questions and statements you make in it are addressed by that hypothesis. What would be more productive, IMO, is responding to those parts of the hypothesis directly, instead of making statements and questions and ignoring the proposed answers to them. In your "levity", you seem to me to have failed to give thos parts any recognition in your reply, and that just doesn't seem useful.

D, sorry about missing your post here initially, but I just saw it, so I didn't want you to consider yourself ignored...

First all of, criticizing someone's inflection, body language, or in this case prose or style of writing, is not a valid criticism of their statements.

Then it's a good thing that wasn't the basis for my criticism of your statements, and only the basis for criticism of your condescension in the absence of support for those statements.

You can't refute something someone says merely by objecting to them putting a or or or a *chuckle* into their writing.

By this you seem to be trying to state that either there was something of substance to refute, or that what I refuted by crticizing *chuckle*s and winks was something of substance. That's puzzling, because the text you quoted of mine illustrates that to be completely divergent from my statements and what you've illustrated so far.

You can try to refute it on that basis of course, but you'll fail. Right? I think you know this.

By this, you seem to be making it your goal to intimate that I failed to refute something you said, by ignoring here my assertion that you didn't provide anything of substance to discuss. The purpose seems to be so that you can then ridicule my commentary on your condescending posting style and be "supported" by the false premise that such commentary was the entirety of the criticism offered.

It's not exactly "levity" that I try to convey as much as it is "lightheartedness"--the idea that while I'm "serious" about what I'm saying I do not however view it as life-threatening.

I don't see how it is a constructive use of time to propose that I'm wrong to use the word "levity" so you can conduct the meaningless exercise of using a different word that carries the same meaning simply as an excuse to further promote the idea that I'm wrong about something without actually demonstrating that the assertion has any validity. What I think you are really doing is dismissing part of the connotation of "levity" by this empty exercise instead of directly addressing it. Here, let me quote a definition of levity, bolding the part you seem to be attempting to circumvent without actually addressing: "1. Lightness of manner or speech, especially when inappropriate; frivolity."

In fact, your entire post up to this point seems a transparent attempt to repeatedly seek opportunities to enforce the idea that I'm "wrong" without actually making a statement in support of that assertion that happens to be true.
Was it to avoid the discomfort of recognizing the initial part of my post that links to your commentary that might serve as validation for Dave H using the term "conspiracy theory" to describe your continued approach to discussing post filters?

I'm sorry you feel as if I've ignored your points--but you seem to be ignoring mine...

Really? Is that why I addressed specific points, and asked you to illustrate your actual hypothesis if you had one? Presumably, this jibe means that you will? Let me annotate the progress of your providing your hypothesis as we continue.

It's great that you'll theorize about the matter of post-filter blending relative to nv3x FSAA, but that's not what I'm after.

Yes, but I was asking for your theory, not for you to take the opportunity to tell me "it's great" that I have a theory but that it isn't good enough for you. Your own viewpoint and opinions require support as well, beyond only stating that another's viewpoint is simply not 100% proven yet. I covered this in the prior post.

Hypothesis: none.

I simply want nVidia to talk about it as opposed to ficticiously maintaining (in my view) that talking about it in general is akin to furnishing the public with circuit diagrams of their processors. I mean, they talk about *so much else* in general that I hardly see any difference here.

Yes, that's what you stated in the post I replied to, and I addressed the validity of that desire IMO....hardly ignoring your points as you assert. However, we're talking specifically about your continued dismissal of a working theory here, without anything more than saying it isn't 100% proven, and while not proposing one of your own, and also the condescension that accompanies that practice.

Hypothesis: none.

It's fine if a hypothesis addresses what I consider to be the issues relevant to me, but I don't want a hypothesis, I want nVidia to talk about the generalities of its use of post-filter blending relative to FSAA in its current products. I'm sure you can see the difference.

Do you really think my understanding of that desire isn't already displayed? Again, this is puzzling, as what you did quote was referring to this directly. Or perhaps you forgot when you didn't include your text of: " I'm just saying it would be nice for nVidia to declassify this "trade secret" (nVidia's words, not mine) and talk about it, explain its benefits, drawbacks if any, etc. That seems an entirely reasonable request. " that you stated just prior to my quoted reply?

Hypothesis: none.

We can promulgate hypotheses from here to Andromeda, but that doesn't change the fact that nVidia's not talking about it, does it? And so none of the hypotheses can be validated or discredited.

Since no hypothesis can be absolutely validated as of yet, you propose that your absence of a hypothesis is just as valid as a hypothesis that fits the observed facts? I think this is what you've been trying to pitch the whole time, and I must recognize the amount of effort you put into hiding how ludicrous the proposition is, but it is still a waste of time, because I already addressed the problem with it despite your attempt to pretend otherwise.

Hypothesis: none.

There's a simple thing anyone who has an nv58/5900 with the latest Dets can do, though, and that's test out the current FSAA modes with the same kind of screen-grabbing software that [H] used in its initial 5800U review--the software that did not capture the post-filter blending. I would do it myself only I don't have the hardware at my disposal.

The beginning of a useful suggestion?

With this screen grabbing software you could contrast shots taken at 0x FSAA with shots taken at QC, 2x and 4x FSAA and inspect them for differences, and then repeat using the screengrab software that captures whatever blending operations are being done in the post filter--then you could determine to what extent if any nVidia is using post filter blending in certain FSAA modes to simulate all, or part of, its FSAA effects.

Well, you do two things...propose a test, and take the opportunity to again propose out of the blue that post filters can only "simulate" FSAA effects (that "if any" is a bit blatant). I've already provided my response to the latter, so addressing the former:

It is too bad you got too caught up in your post filter preconceptions to specify a useful part of such comparison: comparing texture details between the unblended and blended outputs for AA modes. The problem there is that the theory I went through the trouble of proposing for you wouldn't be addressed by that. It is only your persistance in maintaining that your objection-without-hypothesis is just as valid as any hypothesis proposed that would prevent you from realizing that this was stated even in the thread I linked to for the earlier discussion of post filtering.

Making up more hypotheses for you, however: if there was a difference in texture detail, it would show that there was either something unique about the texture information for the two buffers taken as input by the software, or expose a flaw in the "software post filter" that might or might not be representative of a flaw in the hardware post filter.

You might also use this method to determine if post filter blending is being used at all in any of the current products and drivers, and in which FSAA modes the post filter is employed (if it is being used in any of them.)

Actually, what you propose would not show that at all, because you are still caught up in the idea that if it isn't showing AA in the buffer sampled without "software post filtering", that it shows that the final output isn't AA'ed, despite a rather detailed and often repeated explanations as to why that assertion is fallacious. We already have 2x and Quincunx AA modes screenshots that first hand observers say represent the output they see on screen, and if you have observations that contradict that, this would be the time to mention them.
That your viewpoint still reflects this seems to indicate that you are only interested in your anti-postfiltering preconception, and not actually paying any attention to what other people are proposing. I do believe I mentioned time wasting some time ago.

Hypothesis: same one you had in the thread I linked to when illustrating why Dave H used the term "conspiracy theory".

That's the kind of data I'm looking for. My original criticisms of this were ignited by [H]'s original 5800U review a couple of months ago, in which [H] stated it could detect no visual differences between its screen grabs at 0x FSAA and its screen grabs at 2x FSAA, although they could see a difference on screen. This smacks of post filter blending, and the fact that modified screen-grab software was required before the on-screen 2X FSAA IQ could be captured simply underscores the likelihood that this is what is going on (or at least *was* going on at the time.)

OK, did I miss where you explained how your assertions make sense now? Yes, it is post filtering AA. Post filtering does not mean it isn't doing AA.

Those statements are consistent with the observations you just made, and I'm failing to see the mystery in that...unless you just didn't read what people said in response to you when you said this before?

Is it still going on? If so, to what degree and in which FSAA modes?

Ack! Yes post filtering AA is still going on. Did I just miss the reason why you are asking this question?

If nVidia would simply *tell us*, much in the same way it tells us which of its FSAA modes are MS and which are MS & SS, and which used ordered grids as opposed to rotated grids, etc., we'd all *know* and there'd be no need for "suspicions" or "hypotheses," would there?

Edit: typos

Umm...OK, you seem committed to word usage as a 100% substitution for reasoning.
Was there some explanation of why you descended into apparent incoherency about this buried somewhere in your condescension? Again, I'm asking you to point out logic that I may have missed while putting some thought into what has already been said to you, rather than make lengthy posts that completey avoid doing that.

 

[Myrmecophagavir]

Can't we keep this on topic? There was a nugget of graphics-related discussion in your mammoth post Demalion, but it was mostly the sort of pointless waffle you were rallying against shrouding the kernel. Can't you do that in PMs? Yes, I realise the irony of not using a PM for this but it's not really directed at just one person.

 

[demalion]

Myrm, I consider things like that before making posts...what I considered is that WaltC has repeatedly resorted to this type of "argument" for quite a while, and that I was going to address it directly for once.

Did you factor the amount of text WaltC generates by what I am criticizing into your consideration?

I happen to, surprise!, disagree that my post was "pointless waffle" (maybe because you didn't provide any convincing reasons to sway me?), but I do consider it pointless waffle to use an antagonistic label out of annoyance when providing support for statements would be useful...whether those statements are those by myself, WaltC, or, yes, even the critical label you offered. In case you failed to notice, your insult didn't come with any support, just provocation..

Recap: I don't think proposing my reasons are "pointless waffle" without any related reasons of your own added anything remotely useful at all. Do you think it did?

To answer your criticism itself:
If I thought covering the problems I think are in Walt's posts with a PM would have helped, I would have done so. And I have done so with other people when I believe I've already clearly illustrated the reasons for criticism or disagreement...unless, of course, I'm addressed by someone who considers expressing their own annoyance and antagonism more justified than my commentary, because it is their annoyance.

What part of your post was more useful in public rather than in PMs? Did you think I would be convinced to agree with your "pointless waffle" label and so would not reply? You might have perhaps provided me with some reasons beyond it being too "mammoth" for you to read, or done without a label as directly antagonistic and unsupported as "pointless waffle" if returning to the topic was more important to you than expressing your annoyance.

Might I suggest it might have been a more useful alternative to continue the graphics related discussion, and that it doesn't make sense to criticize my post for making exactly that effort because it also made the effort to point out how WaltC is failing to do so?

For myself, I'm not sure if WaltC would use your commentary as some sort of support for his stance. Perhaps he wouldn't, and my posting this outside of PMs to prevent having to de-convolute such an inclusion to address it would then be a mistake. The only thing your decision seems to have accomplished so far is to give me the opportunity to make that mistake, because I just happen to doubt that WaltC would consider your post discouragement for the nature of his statements if my post didn't do the job.
I'm basing that on what I've observed before.

 

[Myrmecophagavir]

Back to the topic at hand, Guru3D reports that HardAvenue interviewed Tero Sarkkinen about Nvidia's response.

http://www.guru3d.com/comments.php?category=1&id=1852

Guru3D has now decided to drop 3DMark from their benchmark suite ! and the news poster has this rather shocking view:

I actually applaud nVIDIA for the way they did it, if you do it then have the b@lls to do it well.

This is the sentence right after he denounces cheating in all its forms!

 

[Doomtrooper]

 

[WaltC]

Then it's a good thing that wasn't the basis for my criticism of your statements, and only the basis for criticism of your condescension in the absence of support for those statements.

D, let's just agree to disagree, OK? Sorry if I was condescending--that was not my intent.

Hopefully, nVidia will one day talk about what it's doing with the post filter in its current FSAA implementations. Then again, maybe it won't. I doubt there's anything anyone could say that would convince me that this is a "trade-secret" topic that deserves the official silent treatment it's getting. But that's just my opinion, of course.

 

[cellarboy]

Did anyone see 'The Screensavers' on TechTV Tuesday night? The whole cheating issue got some TV time. Unfortunately, the headline for the story read 'ATI admits cheating' and Nvidia were mentioned only in the sense that they too had been caught cheating also.....

ATI can't win even when they try to be stand-up guys.....

 

[Reverend]

Did anyone see 'The Screensavers' on TechTV Tuesday night? The whole cheating issue got some TV time. Unfortunately, the headline for the story read 'ATI admits cheating' and Nvidia were mentioned only in the sense that they too had been caught cheating also.....

ATI can't win even when they try to be stand-up guys.....

In Malaysia, we get delayed TechTV programs so I haven't yet seen what you said. But everytime I see the Screensavers program, I see NVIDIA's "The Way It's Meant To Be Played" logos plastered on almost all the PCs used in the program.

Is NVIDIA a sponsor of the program or TechTV in general?

 

[jjayb]

Haven't seen it yet either, just missed it. Will watch it tommorrow. I decided to check out the message board at their website and this was posted by someone there:

On the show when you were talking about cheating drivers, you only mentioned ATI had code in their drivers optimized for 3dMArk. YOu then went on to show screenshots of improperly rendered sections of the benchmark which didnt come from an ATI card but rather from your beloved sponsors Nvidias card.

Please tell me this isn't true. Please tell me they didn't attribute the Nvidia off the rail screenshots to ATI. I'll definitely have to watch the rerun in the morning to find out.

I swear, Nvidia can do no wrong.

 

[cellarboy]

Nvidia sponsor 'The Screensavers' Thursday LAN parties. Kyle B is also their videocard guru too....

They didn't directly atribute the screenshots of the cheating to ATI, but given the gist of the story and the fact they didn't clarify the origin of the shot they did manage to imply that it came from ATI.

Honestly, I was sat listening to the story and couldn't beleive what I was hearing. They took their 'facts' straight from 'The Register' story posted Tuesday morning. Probably mentioned the word Nvidia in the story once.

As they finished, I finally had to admit to myself that Nvidia are going to get away with this scott free. ATI look like villans simply for standing up and admitting their mistake.

Deny, accuse, deny. Obviously the way we should all go in future....

 

[Miksu]

As they finished, I finally had to admit to myself that Nvidia are going to get away with this scott free. ATI look like villans simply for standing up and admitting their mistake.

All we can hope is that B3D or some other site publishes a story from nVidia's cheatings in games (if there is such) so that other sites can't hide behind the "3DMark'03 is just a synthetic test"-wall.

It was rather sad to read Catalyst Maker's response from one of the threads at Rage3D:

wow im lost for words. I dont understand why certain sites still accuse of us cheating. Both Sweeney and Carmack have said what we have done is legit.

Now I have to make sure that code is out of the next CATALYST posting, which I still dont agree with because we did NOTHING wrong. I guess us giving up those 2% and not optimizing in any shape or form is better than giving our awesome end users the best performing part.

Ahhh this is so unfair.

http://www.rage3d.com/board/showthread.php?s=ed548d14a0b639ae1c13bc0ea0a1366d&threadid=33688634