My thread @ Futuremark(Re =Waite for Unwinder)

mboeller said:
Well, a few people in other forums see the Tri-AF implementation of ATi as an cheat as well, cause when you select Tri-AF in the panel you get Tri-AF only on the base texture and Bi-AF on all the other textures. This "inflates" the Tri-AF benchmarks ~30%, as it is said. So for this people ATi cheats on all D3D-benchmarks where the application does not specify AF themself.

That can easily be corrected by either:

a. Setting the in-game and ATi control panel settings differently, or
b. If the game doesn't have it's own settings, then a registry setting change can force trilinear on all texture stages where it's required.

Hopefully this will be made more clear soon, if ATi are indeed planning to change the aniso options in their control panel to include this as a more easily accessible setting.
 
micron said:
Unwinder's should have presented his findings without all the rhetoric, things would have been different if he did. I dont think that it's his work that is under fire here, it's him and his overemotional oppinions.
As one wise man said "I don't agree with any word u say, but I'm ready 2 die for your right to say this".

Think this is "constitutional right" of the author to make conclusions from the fact he have found.
 
RuslanK said:
Think this is "constitutional right" of the author to make conclusions from the fact he have found.
Well, he deffinately has made some conclusions.
Heaven help anyone who doesnt agree with them.
 
RuslanK said:
All
And really I don't know what most of the guys here and there want from Unwinder. He have done a good work and showed his results and his conclusions. If u wnat him to say that vendors act not in exactly similiar way he said this in the article. If u want him to say that ATI's position looks better he said this here.

Well, he did not explicitly state what he said in that forum post in the body of the article he wrote. Now, I think it is common sense that the majority of the people who read the article are not going to ever see or hear of his post in one thread over at Rage3D.

It should be abundantly clear to most people that many readers have short attention spans. Since the conclusion of Unwinder's article is more condemning towards ATI, many readers disregard what was said earlier(that NVIDIA is the worse offender) and come away with the illusion that somehow ATI is at the same level as NVIDIA, which is also not true.

And then we have "gems" like the 3dmark03 Digit-Life article which further go blaming ATI for something they are innocent of.
 
RuslanK said:
Think this is "constitutional right" of the author to make conclusions from the fact he have found.
RuslanK,
We all owe Unwinder a debt of gratitude for taking the time and effort to make the script and making it available. It is another tool to use to weed out the truth. Still, this does not clear up the confusion surrounding his conclusion in the article. Here is a smart guy, he makes this script to uncover some important information. He takes the time o run all these test in a logical manner. But in the end, instead of letting the facts speak for themselves he lets his emotions get the better of him. Sure, he may have been more disappointed in ATI because he had a higher expectation of them. This, however, does not belong in a factual and logical article. Logic and emotions rarely mix. If he was more disappointed with ATI thats his opinion and no one could argue with it. He has a right to it, as does everybody. It’s just that it should have been done in a separate article. By including it in this article some are left scratching their heads wondering about the moral equivalency of the infractions because his own facts show that there is a large discrepancy in the infractions.
 
Re: So surprised by the solid facts, I neglected the conclus

Exxtreme said:
Hi Dave,
DaveBaumann said:
Unwinder said:
I bet that you're talking about NV-styled per-stage AF control. ;)

I'd be surprised if the hardware could do that.
the Cat3.5 has allready this functionality. I know how to switch between bilinear-af, tri-af with stage optimization and "true" tri-af under D3D. :)

Actually it can do even more. I haven't looked in this thread for a while so I hadn't seen the newer d3daf tester version. I've downloaded it and tried it with difference settings on the texture preference slider in the d3d drivers.

Using the slider it seems that the 9800 can actually vary the AF level between the texture stages and also whether it uses AF that is strong at 90 and 45 degrees or af thats strong at only every 90 degress (ie like the 8500)

The settings do the following

high quality: full 16x r300 on all stages
quality: 16x r300 on stage 0, 8x r300 on stage 1, 4x r200 on the other stages
performance: 16x r300 on 0, 4x r200 on 1, 2x r200 on the rest
high performance: 16x r200 on 0, 2x r200 on the rest.

By r300 i mean AF that is max at every 45 degrees and by r200 i mean its max at every 90 degrees like the 8500

Using the application to set linear mipmaps it will use trilinear whatever the tex pref slider setting, using quality AF from the drivers it always uses trilinear on stage 0 and bilinear on the rest.
 
rubank said:
Well said, nelg.

No need to write novels to the same effect.
I hope some people take it as a lesson ;)

My goodness, rubank, IMO that is the most useful criticism you have ever levelled at me! There is no sarcasm in this comment, and I hope any future disagreement with me on your part is directed in the same fashion.

That said, there are specific reasons for my "novels", which I could discuss with you at length (in another location) if you wish. In brief:

I seek to provide a clear body of specific reasoning that a person in disagreement with me can clearly and directly address. However, it is to my dismay that the address of such reasoning rarely, to my perception, occurs, and that rhetorical avoidance of those specifics is a common result. It is, again, my perception, that this latter occurrence is done in the pursuit of being thought by others to be saying something well reasoned (the distinction here is that the actuality of valid reasoning is a lesser detail, secondary to simply having others believe it so), or seen in some other positive fashion or to satisfy some other satisfaction unrelated to reasoning.

For my part, my interest is in actually saying things that are well reasoned, and, in fact, the "novels" and attendant reasoning is what I perceive as necessary to continue to maintain a distinction between my effort and what I describe with dismay just above. This is by the simple expedient of providing that reasoning, in very particular detail, to have others evaluate and demonstrate that it is well done, or not...instead of my (trying to) depend on swaying their evaluation with well turned words, popularity, flattery, intimidation, ridicule, or other similar factors, except as they occur as (distantly, by preference) secondary results of that reasoning, or in direct response to someone else's efforts to use such devices in place of reasoning itself.

But, that's a topic for another thread (in fact, it was the topic of another thread that, to my chagrin, seems to have disappeared, which I might have mentioned before), and I hope that the (relative) length of the above can be allowed if it imparts some information or insight into this topic that was absent before. :)
 
Bambers, that looks to me like a topic for a very informative thread (which might be in progress already...is it?), and I suspect some "Catalyst" personnel might be a bit put out by having the bag so thoroughly empty of the cat. :p It looks a promising direction to take, as long as the issue of application detection with regard to the capability does not become a factor (atleast, not without some measure of disclosure and reasons that clearly benefit users more than benchmark figures, such as, for example, the NwN "leaf" issues).
 
jvd said:
micron
unwinder i just have one question . Why are there other sites using the anti cheat detector that are having much much diffrent results than you ?

Where do you see much much different results? The same huge performance drop in GT4 - Nature after installing AntiDetector.

jvd said:
micron
To me it seems that ati has been honest about every thing but 3dmark2001. Its nvidia that is cheating left and right. So if i was to pick which company to put my support behind I'd pick ati.

Correct. 3DM2K1 seems to be the only application affected by ATIAntiDetector, at least now. And I do hope it is the _last_ ‘optimization’ and I won't be able to find anything more. I do hope that the rest detection routines blocked in the Catalyst are really application workarounds. It will be nice, if ATI will be able to publish full list of applications detected for troubleshooting.

Ichneumon said:
I think the point you miss that is key to all the text that's been layed out in this thread, is that the issue simply is not that ATI is a saint. Instead it is that in comparison to everything seen with the FX series, the sole issue with ATI is one noticed "cheat" (or "optimization" in Nvidia's words) in 3dmark 2001... nothing else. with the FX series there continues to be noted issues with at this point a minimum of a handful of regular benchmarks, and more keep being brought to light every day.

Invasion from Rage3D? :) Nope, I absolutely don't miss this point, even before posting the article I mentioned on the forums that Detonator FX suffered from AntiDetector much more then Catalyst and I perfectly understand that NVIDIA definitively went way *too* far with their Detonator FX ‘optimizations’. They will get what they’ve deserved too, don’t worry. The article clearly states that FX GPU family get really serious performance hit from AntiDetector almost in all D3D benchmarks and states that it will be investigated in upcoming FX5900 reviews. BTW Russian version of Gainward FX Powerpack Ultra/1200 Golden Sample (FX5900) review containing NV/ATIAntiDetector benchmarks with shameful (for NV, not for ATI) CodeCreatures and UT2003 benchmarks and NV-crushing conclusions is already launched: http://www.ixbt.com/video2/gffx-16.shtml
So it is definitively not a missed point. Furthermore, after launching 3DMark2001 related article ATI ensured me that I'll hardly find anything but 3DMark2001 related tricks inside Catalyst and I do hope and _believe_ that this is the real truth and not a PR.
But look at the article from different point of view. This "nothing else" you are talking about is extremely popular synthetic benchmark. Nature is a part of this benchmark, this subtest was used as de-facto for estimating PS/VS performance during long period of time and it is still used in a lot of reviews. Do you like to understand that Nature optimizations were first introduced by ATI after R200 launch? I don’t. Do you like to understand that we’ve never seen the real R3x0 performance in Nature - GT4 before? I don’t again. Can we use these facts to rehabilitate NVIDIA and tell that boost in GT4 in Detonator 40.xx is just an answer to ATI tricks? Of course we cannot. Do you like to anderstand that 'ideal' ATI is able to cheat too when they need it? There are no excuses for cheating, there are no reasons allowing legitimate it. There are no good and bad cheaters. Either you cheat or you don’t cheat, for me there are no alternatives more. Attempts to rehabilitate vendor just because the competitors act worse have no sense. It’s the same like to say ‘we can simply forgive a killer just because there are a lot of bad guys like Saddam who killed much more people’. Both are guilty now, whatever you like it or not.
The most important thing we’ve to look at is _reaction_ of both vendors on this publication. Both are busted, but both can make step back. I don't hope that NV will do it, but I do hope that ATI will make correct conclusions.
There is no way to change now days, there is no way to change the past and the current situation will not look better (however it can become ever worse in case of new tricks detection) but it is possible (or at least we can try) to change the future.

Ichneumon said:
Your rhetoric in various threads i've seen goes contrary to the very conclusion that you have been defending in this thread.

I don’t see contrasts between my postings, but I’ll be pleased and try to address them if you’ll find and show them to me. I’ve made some public statements related to AntiDetector and I’m still ready to defend each of them:

1) Both leading hardware vendors ‘optimizie’ drivers for benchmarks long ago and unfortunately 3Dmark2003 scandal is nothing but the continuation of old old ‘optimization’ story. Sad but true. Regardless of PR claims posted by each IHV, benchmark specific ‘optimizations’ are still inside the drivers.

2) ATI’s post-investigation reaction definitively inspires respect. NVIDIA’s reaction with 44.67, contrariwise shames them.

Ichneumon said:
If 3dmark 2001 were of any consequence whatsoever today, perhaps that could be argued to be some major sin... and in that case, I expect ATI would remove the optimization just like it did the one noted in 3dmark2003. Perhaps in the next driver release we'll see that happen, or sometime soon when something of such trivial impact doesn't take time away from other legitimate driver work.

I hope.

Ichneumon said:
On the other hand, it is as if there can't be testing on enough apps to keep up with Nvidia's cheats both in benchmarks and in regular timedemos used to bench games. There simply is no comparison between what has been seen in the unprecidented amount of cheats to prop up the failures in the FX architecture, and one lone leftover optimzation for an effectively obsolete benchmark (for a r300 class graphics card) in ATI's current WHQL released driver.

I really wish I can perform such testing myself. Unfortunately the most of DetFX’s optimizations are NV3x specific and the only application affected on my Ti4600 is 3DMark2001. Anyway, NVAntiDetector are walking in the net and a lot of reviewers already started testing FX’s with it. So we just have to wait and see.

micron said:
Yes, I have more to say.
When people see you in forums, and ask you questions about your Rivatuner tool, you should try not to be such an A-Hole to them, or snap at them, or say things like "what...you cant read!" or "I'm not going to help you if you cant read directions"
These people your 'snubbing' could in a way be called your fans, they think your pretty knowledgable, but your attitude shows how you really feel about most forum members. I didnt see all this at first until another B3D member pointed it out to me.
You've got an attitude that kind of screws up the cool things you are doing. The minute you think that someones disagree's with you, you throw a small fit.......too bad....

??? Are there any relations between RT and my forums and the article being discussed here? Please add a bit logic and motivate your insults or simply step back.
Towards to your claims, it is fun that you're judging RT forum without ever visiting them before. If a visitor is lazy enough to use search or he simply ignores built-in context help and posts the question included in built-in FAQ, he will get only one reply: RTFM. If it means being A-Hole for you then you've simply never coded and supported something really huge. RT is 2.000.000+ users tool, simply try to support forum in different maner then I'll look at you.
Nevermind, offtopic addressed to your claims and insults is finished. My apologies to moderators for discussing RT in this thread.
 
Unwinder first i'd like to thank you for responding to my post here in a much better ton and more insight full way than over at rage3d.

The second thing I want to ask is how do we know this isn't simply a work around for a bug in say the 8500 that ati put in to fix a problem and it inflates the r300 scores. Much like the quack fiasco. I really think before we can start throwing stones we need more benchmarks of older drivers to see really where this scandel started. If we can track it down to the first cheats in each driver all we need to do is figure out which driver came out first.
 
Unwinder, you go girl. :LOL:

I agree with what you have said. So I guess I can say I second that. :)

Jvd: Unwinder's comments are very solid. He doesn't mouth off at things he knows very little or nothing about. :)
 
K.I.L.E.R said:
Unwinder, you go girl. :LOL:

I agree with what you have said. So I guess I can say I second that. :)

Jvd: Unwinder's comments are very solid. He doesn't mouth off at things he knows very little or nothing about. :)

I'm not saying he ever did . I'm saying the way he sometimes says things and how brief he is with his respones are not apreciated. I'm not pretending i know everything about this but if he took some time and explained more indepth in his respones it would help.
 
jvd said:
The second thing I want to ask is how do we know this isn't simply a work around for a bug in say the 8500 that ati put in to fix a problem and it inflates the r300 scores.

Please read the article carefully. First, R200 and R300 d3d drivers are loacted in different files so R200 workarounds simply cannot affect R300+ score. Second, different approaches are used to 'boost' Nature in R200 and R300. R200 driver uses vertex shaders to detect the scene whilst R300+ detects the scene via texture creation.

jvd said:
I really think before we can start throwing stones we need more benchmarks of older drivers to see really where this scandel started. If we can track it down to the first cheats in each driver all we need to do is figure out which driver came out first.

I doubt that it's possible to find cheats daddy. The story started really long time ago. The first ATI trick I can remember is 3D WinBench detection and 40% boost in Rage Pro drivers. I've no information about NV cheats at that time, but I'm pretty sure that if I'll dig Riva128 drivers I'll also find something like that there.
Once again: the most interesting thing for me is not 'who started this game and who is better'. The main point is 'who'll stop this shame at last'.
 
I didnt even know there was a Rivatuner forum :D
Those were'nt the forums I was talking about!....oh well ;)
 
Unwinder said:
Invasion from Rage3D? :)

hehe, not quite... i've been readin beyond3d since before I ever worked for Rage3D... and that's quite a long time. :)

I wanted to thank you for your well layed out responses to my post. There are a couple things we don't entirely agree on, but I definately understand where you are coming from a lot better now.

I will be inquiring about the status of the 3dmark2001 gt4 optimization to see if there is anything in the works about addressing that from ATI.
 
Unwinder said:
jvd said:
The second thing I want to ask is how do we know this isn't simply a work around for a bug in say the 8500 that ati put in to fix a problem and it inflates the r300 scores.

Please read the article carefully. First, R200 and R300 d3d drivers are loacted in different files so R200 workarounds simply cannot affect R300+ score. Second, different approaches are used to 'boost' Nature in R200 and R300. R200 driver uses vertex shaders to detect the scene whilst R300+ detects the scene via texture creation.

jvd said:
I really think before we can start throwing stones we need more benchmarks of older drivers to see really where this scandel started. If we can track it down to the first cheats in each driver all we need to do is figure out which driver came out first.

I doubt that it's possible to find cheats daddy. The story started really long time ago. The first ATI trick I can remember is 3D WinBench detection and 40% boost in Rage Pro drivers. I've no information about NV cheats at that time, but I'm pretty sure that if I'll dig Riva128 drivers I'll also find something like that there.
Once again: the most interesting thing for me is not 'who started this game and who is better'. The main point is 'who'll stop this shame at last'.

thanks for the info. It seems at least ati has started cleaning up the pile of dog poo . Hopefully nvidia will join soon but it seems they are just rolling around in it .
 
demalion said:
Bambers, that looks to me like a topic for a very informative thread (which might be in progress already...is it?), and I suspect some "Catalyst" personnel might be a bit put out by having the bag so thoroughly empty of the cat. :p

I wasn't intending on doing a thread, i might put my 8500 back in and have a look at that with the different stages.
 
Unwinder said:
But look at the article from different point of view. This "nothing else" you are talking about is extremely popular synthetic benchmark. Nature is a part of this benchmark, this subtest was used as de-facto for estimating PS/VS performance during long period of time and it is still used in a lot of reviews. Do you like to understand that Nature optimizations were first introduced by ATI after R200 launch? I don’t. Do you like to understand that we’ve never seen the real R3x0 performance in Nature - GT4 before? I don’t again. Can we use these facts to rehabilitate NVIDIA and tell that boost in GT4 in Detonator 40.xx is just an answer to ATI tricks? Of course we cannot. Do you like to anderstand that 'ideal' ATI is able to cheat too when they need it? There are no excuses for cheating, there are no reasons allowing legitimate it.

This is a piece of pure conjecture on my part, but I would imagine that if ATi had been the first to cheat in 3DMark 2001, we would probably have seen nVidia pick up on it and offer that information to the online press in the same way they did with the whole Quack saga. If they'd have done that and been able to come up smelling of roses they would have done so, because it would probably have all but finished ATi and the 8500s reputation at that time.

At the end of the day of course it doesn't matter 'who started it', cheating because someone else did is no excuse, but I thought I'd throw my thoughts in anyway.
 
Back
Top