PDA

View Full Version : Nvidia is quacking up, 3dmurk2k3 confirms it


Pages : [1] 2 3 4 5 6 7 8 9 10 11 12 13 14

Skynet
06-06-03, 12:17 AM
here we go again.




http://www.tech-report.com/etc/2003q2/3dmurk03/index.x?pg=1in a funny twist of fate, I got a tip earlier this week about NVIDIA's Detonator FX drivers. The allegation: if you rename 3DMark03.exe to something else and run the benchmark with anisotropic filtering enabled in the drivers, test scores drop. In other words, NVIDIA appears to be using the same lame technique ATI did way back when: keying on the program's filename in order to trigger benchmark "optimizations." In this case, those optimizations appear to be a lower quality form of texture filtering than the anisotropic filtering method selected in the driver control panel. Many review sites like us benchmark cards with anisotropic filtering and edge antialiasing turned on, so these things do matter.

Hellbinder
06-06-03, 12:41 AM
i would start testing this with as many games as possible. I commented in another thread here today that I thought Nvidia was still Doing some Fishy things with AF... Being that there was no reason for their big AF gains, especially since they only appear to be in certain major benchmarks...

Edit: there are other ways to do this kind of thing without this Kind of application detection. I personally believe they are pulling the same kinds of things in

Serious Sam
Quake3
Unreal Tournament

SlyBoots
06-06-03, 12:42 AM
some have already asked> "just how deep is the rabbit hole?":confused:

SlyBoots
06-06-03, 12:46 AM
Originally posted by Hellbinder
i would start testing this with as many games as possible. I commented in another thread here today that I thought Nvidia was still Doing some Fishy things with AF... Being that there was no reason for their big AF gains, especially since they only appear to be in certain major benchmarks...

I noticed this comment in another forum> "nVidia also broke the getpixel function for opengl in their 2 lowest AA modes"

an off the cuff remark with no follow-up......so, what does it mean??:confused:

Hellbinder
06-06-03, 12:51 AM
You also have to wonder...

Didnt Nvidia realize with everything else that something like this would get uncovered??? I cant believe they went this far, considering everything else.... Its pretty damn scarry.

Oh and Nv40.. I Guess this negates that thread you made about Nvidia Wining 3dmark with AA+AF evern without the *optimizations*.

gordon151
06-06-03, 12:55 AM
Originally posted by Hellbinder
Oh and Nv40.. I Guess this negates that thread you made about Nvidia Wining 3dmark with AA+AF evern without the *optimizations*.

Hehe that was one of the first few things that came at me after I read the article and came back in here. His thread about it is just a couple steps down too :).

Hellbinder
06-06-03, 12:59 AM
With all the recent Futuremark-Nvidia negotiations and agreements going on behind the scenes. And Official Statements of not cheating and getting special IHV support and the other things suggested in the Nvidia-Futuremark statements.... and yet...

All the While Nvidia knew damn well they were Still doing something like this??? While seemingly threatening leagal action??

Its just unbelievable.

gstanford
06-06-03, 01:09 AM
Has anyone bothered to do a diff of the two images against the reference rasteriser?

You never know - nVidia's new anisotropic could be both faster and more correct all at the same time...

Hellbinder
06-06-03, 01:11 AM
Has anyone bothered to do a diff of the two images against the reference rasteriser?

You never know - nVidia's new anisotropic could be both faster and more correct all at the same time...

Hello... It only goes up with a specific application detection. not some general improvement in the rendering process. This is not some *new* AF process..
:rolleyes:

ChrisW
06-06-03, 01:13 AM
I really find it hard to believe that nVidia would detect the name of the application after what happened to ATI for doing the same thing. :confused:

Skynet
06-06-03, 01:15 AM
Has anyone bothered to do a diff of the two images against the reference rasteriser?
You never know - nVidia's new anisotropic could be both faster and more correct all at the same time... That's what I call blind faith.

This is the biggest fiasco I have ever seen in the hardware biz. What did Nvidia do, hire a whole team of driver cheating coders? *rubs eyes*
My brain keeps telling me that Nvidia would not be that stupid as to pull a quack like move. There is some mistake, right?

Behemoth
06-06-03, 01:16 AM
what the hell... unbelievable!!!!!!
nvidia is really crossing my bottom line now...

gstanford
06-06-03, 01:20 AM
For the moment I think the renaming part is coincidental.

The real issue is: is the anisotropic filtering better or worse than before?

If it is better, you have nothing to complain about.

ChrisW
06-06-03, 01:21 AM
I would like to see someone rename Hypersnap and compare snapshots taken from it. I've read too many posts from GFFX owners that state snapshots taken with Hypersnap look nothing like what they see on the screen.

Behemoth
06-06-03, 01:23 AM
3DMark looks quite better than 3DMurk to me. which means the "cheat" is trying to make things look better and faster.

extreme_dB
06-06-03, 01:28 AM
I don't understand why Nvidia puts SO much importance on being the fastest that they have to pull off all these underhanded tricks. Why not emphasize quality for once, and remain focused on all the other aspects like driver stability that have helped establish their dominance?

It's becoming a mad house! :eek:

Hellbinder
06-06-03, 01:28 AM
For the moment I think the renaming part is coincidental.

The real issue is: is the anisotropic filtering better or worse than before?

If it is better, you have nothing to complain about.

Are you seriously trying to suggest that the scores went up by like almost 1,000 points with AF on becuase they *improved* the Quality???

Now i have read it all.

digitalwanderer
06-06-03, 01:29 AM
Originally posted by gstanford
For the moment I think the renaming part is coincidental.

The real issue is: is the anisotropic filtering better or worse than before?

If it is better, you have nothing to complain about.
No, if it's application specific it IS something to bitch about...if it was just a new form of anisotropic filtering it would be enabled in everything. :rolleyes:

I'm done saying "I can't believe they'd do this", the evidence keeps overwhelmingly proving me wrong. :(

At least we know why they take so long between driver releases lately, and why they've been getting so much bigger. :lol:

digitalwanderer
06-06-03, 01:31 AM
Originally posted by Behemoth
3DMark looks quite better than 3DMurk to me. which means the "cheat" is trying to make things look better and faster.
You're REALLY reaching now... :lol:

DSC
06-06-03, 01:31 AM
Tech Report didn't test any ATI cards? Now you can easily guess who passed this on to them.... revenge motivated because of Quack?

I wanna see the scores on ATI cards on 3dmurk2k3, should be interesting since they were also caught doing funny thing with patch 330.

digitalwanderer
06-06-03, 01:35 AM
Originally posted by DSC
Tech Report didn't test any ATI cards? Now you can easily guess who passed this on to them.... revenge motivated because of Quack?
Did you even bother to read the article? You don't have to guess, they specifically state that it was ATi who tipped them off to it after they had been analyzing the latest detonators for "surprises"

What, you don't LIKE the idea that ATi busts nVidia? Tough! :lol:

StealthHawk
06-06-03, 01:36 AM
Originally posted by gstanford
For the moment I think the renaming part is coincidental.

The real issue is: is the anisotropic filtering better or worse than before?

If it is better, you have nothing to complain about.

It's worse than before. The first anisotropic filtering speed up came in 43.51. This is the same speed up that was seen in reviews with 44.03. And quality in 44.03 is degraded.

Discussion (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=11672&pagenumber=3)

You can clearly see that the sampling pattern of Quality AF got worse. Who knows what else nvidia is doing under the hood though. Whether they are doing something else legit or just cheating again to further increase performance, that remains to be seen. Application detection is a poor way of optimizing AF though, no matter how you look at it.

ChrisW
06-06-03, 01:38 AM
Since they used build 330, I guess they didn't find all the 'optimizations'. What else are they going to find?

Skynet
06-06-03, 01:44 AM
What else are they going to find?
They are going to uncover the whole truth: Nvidia actually has built in a satellite transceiver in every FX card that detects though the drivers what benchmark is being run, and the pre-rendered images are streamed to the video card from Nvidia headquarters in real time. It will be revealed that the FX chip is actually nothing more than an smooth block of silicon with no switching logic at all. The circuitry around it is actually the leading edge in wireless datastreaming technology.

:rolleyes:

DSC
06-06-03, 01:47 AM
digitalwanderer,

DOH! My apologies, I skipped the first page and went straight to see the scores and the images.... :o