UT2K3 Filtering AMDMB.com tell it like it is

[Randell]

http://www.amdmb.com/article-display.php?ArticleID=247&PageID=1

Nice investigative work.

Not from a site you'd expect to delve deep into graphic IQ issues either.

 

[Himself]

Cool

 

[MuFu]

So in 44.03 "Performance" mode, AF is bust completely in UT2.003K? Did [T] pick up on this, I can't remember?

MuFu.

P.S. I bet "Mr Hung" gets all kinds if interesting offers. 8)

 

[Natoma]

Hmmm. This is really interesting. Sigh. I can't believe that people are so blind that they're willing to actually ignore the truth and parrot these obvious falsehoods. Nvidia's obviously not doing the correct filtering. Thank goodness for B3D.

p.s.: Mufu, I'm assuming you're referring to HardOCP with the [T] comment, mostly because of the [H] logo. But why a [T]?
What's the joke?

 

[Slides]

[T]ardOCP as some people like to call it. Personally, I don't thing that's needed.

Anyway, this is an excellent article that tells it how it is, without drawing any undue conclusions. It clearly lays out the facts, without any underlying spin and lets the reader decide. I like it.

 

[WaltC]

Very good article actually which might have been entitled "Here's what [H] isn't telling you"....

I thought the opening excerpts from the nVidia driver promotional info were very telling as to the real position of the nVidia corporation with respect to any and all benchmarks--as the "30% performance improvement" claim was based *entirely* on one of the canned timedemos that ships with the game UT2K3--Antalus Fly-By, which is not only a benchmark--it's not indicative of actual UT2K3 gameplay at all (as the bot-match benchmarks are.)

I feel sorry for [H]--while they are doing their best to support some sort of imaginary "nVidia cares not for benchmarks, but for 3D games, instead" position, nVidia is busy making fools of them by publishing so-called "performance increases" for driver sets that are 100%-benchmark based. The fact that the performance increases are clearly achieved by sacrificing image quality simply adds insult to injury for [H.]

 

[Randell]

[T]urd I suppose

 

[MuFu]

The conciseness of the article is excellent and I like the way he concludes it - very non-commital.

Has somebody been able to reproduce these screenshots? I suppose this "bug" affects the entire NV3x lineup.

To be honest I don't think this is as much of an issue as the filtering problem. When people select "Performance" they expect to take an IQ hit, albeit a relatively small one. The same can't be said for "Quality" mode where you really expect full trilinear etc. The Way It's Meant To Be Played and all that jazz...

MuFu.

 

[martrox]

Hmmm.....I wonder if Mr. Bennett is going to now pull his support from AMDMB concerning this:

http://www.athlonmb.com/article-display.php?ArticleID=243

it would fit the way he seems to run his forum......

 

[TheMightyPuck]

It goes up to 11.

 

[Jerky]

Hi everyone.

I'm the author of the article at AMDMB. Just dropping in to answer some questions raised here.

I'm not sure if people saw the minor but significant update to my article, but there was a typo. The setting that was posted earlier afternoon was "Performance" while in reality it was supposed to read "Quality".

All my tests were run using the highest possible settings so I can have the best fidelity in image quality.

I encourage people to try to reproduce the problem. It's as easy as:

• Reinstalling v43.45 drivers
• Turn on AF8.
• Run UT2K3. Record the FPS.
• Take a screen capture using the CTF-Anubis benchmark batch file.
  
• Reinstall v44.03 drivers
• Turn on AF8.
• Run UT2K3. Record the FPS.
• Take a screen capture using the CTF-Anubis benchmark batch file.
  
• Compare images and FPS results from 43.45 and 44.03

I'm glad people here found the article informative. I've tried very hard to produce something that was unbiased and credible.

MuFu:
You'd figure with a last name "Hung" I'd get a lot of interesting encounters, but I don't. If you know of any nice girls who like that name, tell them to contact me... and don't let my wife know!

Thanks for reading everyone!

- Jonathan.

 

[Archaeolept]

some typo there Jonathan

It sounded to me like just more of the same nVidia Marketing meets driver BS, but I really don't expect anything from "performance" anisotropic.

That it was under "quality" settings makes this much more interesting; and, in fact, seems such an extreme failing that I would give creedence to it being just a coincidental bug (even given nVidia's poor history of opportunistic bugs) if it hadn't been specifically singled out by nVidia marketing. That's just too coincidental for me - surely any driver team worth half it's salt would notice that only in the most used benchmark were there completely unexpected and massive gains. Claiming this front and center in their driver marketing PR makes it pretty difficult for nVidia to claim ignorance/"shucks, how did that get there".

 

[Jerky]

Yeah, that typo was pretty bad. A forum member at AMDMB asked me: "Why are you running Performance setting?" and I was thinking, "How did he come up with that? I'm running Quality!".

After checking what was released online, I quickly got my editor on ICQ and had it corrected.

Well, NVIDIA is now in the hard position of improving texturing while trying to maintain frame rates (since customers are now accustomed to the performance levels).

I certainly do not want to be part of their driver team.

- Jonathan.

 

[Seiko]

The conciseness of the article is excellent and I like the way he concludes it - very non-commital.

Has somebody been able to reproduce these screenshots? I suppose this "bug" affects the entire NV3x lineup.

To be honest I don't think this is as much of an issue as the filtering problem. When people select "Performance" they expect to take an IQ hit, albeit a relatively small one. The same can't be said for "Quality" mode where you really expect full trilinear etc. The Way It's Meant To Be Played and all that jazz...

MuFu.

Yes unfortunately one good article can't turn the growing tide. Now it's time for many other sites to run their own tests and publish the findings.
It's only when the review sites unite in a common theme will the big IHVs begin to takle note. At the moment each unethical optimisation is only covered by 1 or 2 sites at any one time. Now imagine if every review site hit out when discoveries like this where found. The IHVs would no doubt think very differently. Sure some sites would have a little egg on their face but lets be honest wouldn't we all respect them a little more for acknowledging they where duped?
So in short and I've said this many times, only when the sites formally agree to run a story at very similiar times will the IHVs stop and take note. Until then the IHVs realise the sites are nothing more than marketing extensions. I guess the site owners need to determine how they wish to be percieved. Creditable journalist or bedroom website author?

So can I please start the ball rolling and ask Dave and co. to set some time aside asap to check this story out and publish the findings? After all I don't think Dave runs this site from his bedroom does he

 

[Reverend]

It's a little disappointing to see that neither AMDMB nor [H] gave credit to who discovered this first. Oh well, what can you expect.

 

[Seiko]

nah, you're bigger than that!

It's a little disappointing to see that neither AMDMB nor [H] gave credit to who discovered this first. Oh well, what can you expect.

I'm sure you're bigger than that, after all isn't better to be part of the machine that stops this kind of unethical optimisation? It will surface soon enough the leading authorities but it's better to at least group together.

I hope to see B3Ds headlines soon?

 

[MuFu]

nV seems to be falling into this cycle of releasing a driver that includes a hack so it can boast "30% performance improvement" then getting busted, removing the hack and just introducing a new one so they can boast more performance increases in the next release. Quite bizarre, although I expect it will continue until the fall when we might see some substantially improved hardware.

Their "adaptive" filtering technique sure is clever... it can even detect which site is authoring an article!!!

Thank you for dropping in and correcting the typo, Jonathan. That certainly makes things more interesting.

MuFu.

 

[Reverend]

Re: nah, you're bigger than that!

It's a little disappointing to see that neither AMDMB nor [H] gave credit to who discovered this first. Oh well, what can you expect.

I'm sure you're bigger than that, after all isn't better to be part of the machine that stops this kind of unethical optimisation? It will surface soon enough the leading authorities but it's better to at least group together.

I hope to see B3Ds headlines soon?

Oh no, don't get me wrong -- all I want is for all IHVs to learn the lesson that there are individuals/websites out there that will eventually discover these sort of things. The thing is that as things progress (for the worse, as it seems, due to NV's continuous tactics) it is best that things are set straight, as in who/which website is ultimately the source of certain discoveries. Neither me nor Dave nor B3D as a site is after any sort of "glory" -- Dave had decided that B3D is "rather too small a fry" to break the news with this UT2003 NV filtering trickery and that it would be more interesting to see how things pan out if some "bigger" website breaks this (with the way they break this news) -- the results as seen from [H] and the other websites regarding this has been, well, interesting... we know more about websites now, don't we?

 

[Jerky]

It's a little disappointing to see that neither AMDMB nor [H] gave credit to who discovered this first. Oh well, what can you expect.

Hey Reverend,

Believe me, if I had known about Dave Baumann's thread while I was conducting my experiment I would have given proper recognition. At the time, we did not realize Dave had started a similar thread here in your forums discussing the issue.

My motivation came from some irregular data points gathered while writing another article on AMDMB. Ryan and I decided to write a second article investigating the issue.

I know you don't know me from some other guy, but trust me when I say that if I had known about Dave's thread here in your forums, I would have given proper recognition to it.

However with that said, I'll talk to my editor and see what I can do in getting props in there for you.

- Jonathan.

 

[Dave Baumann]

Hey Johnathan - Really, don't worry about it. You've taken things from a more AF perspective, which quite frankly I've not even begun to dig into the effects of yet. We've been looking at this from the pure non-AF Tri/Bilinear filtering perspective so far.

Its certainly good to see a number of sites digging into these issues though, and getting a good range of opinions and analysis. At least more people (reviewers and consumers) are being more aware these days.