3DMurk03 new cheats ?

[radar1200gs]

Please head over here:
http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=12996

download the images on offer and comment please.

(this is a test using NOLF.exe renamed to 3DMark03.exe for the lazy - NOLF is a DirectX title).

Could someone please provide hosting for 3 high quality jpeg images at approx 370kb apiece please? (images are low quality because of formum space restrictions).

 

[Dave Baumann]

Greg, The AF issue occurs when AF is selected in the app, not the driver control panel settings.

 

[radar1200gs]

Thanks for the reply dave.

I am specifically adrressing two things here:

1) That the drivers cheat based on an executable name.

2) Those people who claim that because nVidia are cheating in 3dmark2003 they are also cheating in other games as well.

 

[radar1200gs]

NOLF 2 has an anisotropy toggle in it, so in the case of NOLF 2, the app would be selecting it.

I own NOLF 2 and it would be no trouble to run an identical test to the NOLF test I just performed, if you would like.

 

[Dave Baumann]

OK, here are some shots from the 3DMark texture filtering tests (2.6MB):

http://www.beyond3d.com/misc/3dmrendering/3dmurk.zip

If you compare the 1x 3dmurk to the 8x 3dmurk shots you'll not that the quality of filtering improves through all the texture sections. Contrast that with the 3Dmark 1x and 8x you see that only every other texture section gets improved filtering. With the name as '3DMark03' the red/green pink/yellow secions always appear to stay at the same filtering level as 1x, while the black/white and blue/pink sections do see improved texture filtering.

 

[RussSchultz]

Wasn't that method one of their performance anisotropic settings?

Oh, wait. Never mind, that was only doing aniso on the first stage.

Hmmm, that might have something to do with it, but probably not.

 

[Dave Baumann]

I don't think any of theor performance settings said to apply AF on alternate textures on screen when they are displayed on the same plane!!

 

[radar1200gs]

Dave, I honestly could not care less what does or does not happen inside 3DMark2003 itself. It is totally irrelevant to anything, always has been and always will be.

I might also add that you have no proof that it is the driver doing the degradation and not futuremarks patch. Have you tried your tests under the original 3DMark2003?

The real issues for me are the allegations of cheating by executable name (we all know who started that little cheat rolling, don't we?) and the allegation that the driver uses the same sort of cheats on actual games. None of the rest of it interests me in the slightest.

 

[RussSchultz]

I don't think any of theor performance settings said to apply AF on alternate textures on screen when they are displayed on the same plane!!

Well, if its set up where each alternating texture is in either stage 1 or stage 2, then it sort of makes sense.

Or not. Just throwing possibilities out there for people to mull over.

 

[Dave Baumann]

Greg, What you may feel is irrelavent is entirely beside the point in relation to this thread, since the thread is discussing the Tech-Reports article to do with these Anisotropic filtering issues - if you feel that this is entirely irrelavent then don't participate in the thread. The screenshots I've posted are 100% relavent within the context of thread.

As for the patch issue, yes, that was with the 3.2.0 patch. Suggesting that Futuremark have intentionally done something is a little absurd don't you think? These results are reproducable by anyone and I'd encourage you to do so.

 

[epicstruggle]

Wow, must remind my self never to piss off Wavey.

Anyways, are the anomilies (or whatever you want to call them) only happening to the current drivers? If someone could test (and post screenshots and results) using the old drivers prior to the performance boost, this would greatly be appreciated.

later,

 

[madshi]

I honestly could not care less what does or does not happen inside 3DMark2003 itself. It is totally irrelevant to anything, always has been and always will be.

You should better say "I find it totally irrelevant, I always did and I always will". After all you're just stating your personal opinion, right?

 

[radar1200gs]

Dave, how did you capture the 3dmark screenshots please?

I've downloaded your file, and you have to have something set wrong somewhere I'm afraid. Thhe image quality is terrible compared with what I am seeing. Are you sure the capture utility is giving you a true represntation of what is on screen?

 

[Dave Baumann]

Press F12.

FYI: This is with 'Application' AF setting (i.e. off) in the driver control panel and AF selected via 3DMark.

Also, if you change the texture map to black and white for the entire scene you'll note that there are changes across the entire thing (I'd assume the same texture will be in use here), its only when you change to the coloured textures that it appears to be a little more selective in where the AF is applied.

 

[radar1200gs]

I get identical screen shots for murk and mark using your method.

In fact the screenshots are indentical to setting the level of anisotropy in the driver panel.

I can provide the .bmps if you want.

bear in mid I have a GF3, not a GF-FX.

 

[Dave Baumann]

The previous shots were with a 5600. The same thing does not occur with a GeForce 4 either.

 

[radar1200gs]

The only conclusion I can reach is that if there is an anisotropic filtering anomoly it is in the actual NV3x silicon, and not the detonator drivers.

We all know NV3x did not go according to plan at the foundry. It has been said that anisotropic filtering and a lot of FSAA modes were broken on the silicon used at the NV30 launch in september. My guess is this is probably a bug/fault that never was entirely fixed.

 

[Dave Baumann]

The silicon is coded to recognise "3DMark03.exe"?

 

[John Reynolds]

The silicon is coded to recognise "3DMark03.exe"?

Ahh, that accounts for the transistor increase of NV35 over NV30.

 

[radar1200gs]

No, the silicon is buggy.

You have access to GF-FX's; why not try out the two tests I have put forward (the aniso app and the NOLF test) on them?

They will either exhibit the bug or they won't and nobody will know for sure until the tests are done. I know for sure that nothing untoward happens on GF3 level hardware.

If I had access to GF-FX hardware myself, I would personally run the tests, but I don't so I have to rely on others to do it for me. So far nobody seems very willing.