On 3DMark Issues

[Dave Baumann]

Sorry for bringing up 3DMark again, but those with NVIDIA boards may like to try something. Create a shortcut to 3DMark03 and then put the options "-nosplash -noscreens" in the shortcut. You might find a small performance difference to running it without these options (about 1 or 2 FPS or so in some tests).

 

[Slides]

Interesting. Please elaborate if you can.

 

[FUDie]

Interesting. Please elaborate if you can.

Last year we had "Splash Screens 2001". This year we have "Splash Screens 2003". I guess nvidia didn't learn their lesson.

-FUDie

 

[K.I.L.E.R]

I'm going to screem foul and call it a cheat.

 

[Slides]

Err ok, but what exactly is it?

 

[Randell]

Err ok, but what exactly is it?

based on last years evidence it would appear that nVidia's 3DMark2001 optmisation routine detects the splash screens to trigger it. Maybe 3dMark03 does something similar as well.

No need to scream cheat as there is no proof that last years routines degraded IQ.

 

[tertsi]

It have affect only on GT1 (Wings of Fury), like 3-5fps. Tested with 5800 Ultra and 44.03's.

 

[Dave Baumann]

I get minor differences through all tests on the NV30.

However, If I run this through the developer version then this clears one of the issues. The effect that is most notable in the first screenshots is cleared up.

 

[demalion]

Can you quit 3dmark before a game scene starts being drawn but after a splash screen is shown? If the detection routine doesn't have a time elapse or some other further check for intialization, it would be interesting if behavior for some other D3D application could be checked.
Seems pretty unlikely for detection to have that failing, though.

 

[Himself]

Guess those command line options don't work with the free version..

 

[John Reynolds]

No need to scream cheat as there is no proof that last years routines degraded IQ.

That's right! If you can't see it, it's an optimization and not a cheat!

 

[demalion]

...
By admitting those two new partners, FutureMark gets two very convenient outlets when they wish to issue slaps within the industry. This is something they have never had the ability to do. Nice huh? Convenient too, no? Certainly interesting. Assassins for hire?
...

Something at 3DChipset that reminds me of Kyle's rant. I've decided not to linkify, because I have some trouble with the idea of rewarding sensationalism based on innuendo with ad revenue. Conversely, I didn't quote the whole thing, because that would be going too far in the other direction (IMO).

Of course, people are entitled to their opinions, but how do you correlate a writer keeping trust with their readership with publishing opinions with "support" based entirely on innuendo? Or is that priority lost somewhere in offering an "opinion"?
"Solomon" provides a differing opinion, but even with that the piece still manages to completely avoid mentioning the remotest possibility of blame for nVidia...Beyond3D, ExtremeTech, and Futuremark are the entire focus after mentioning 'the NVIDIA "cheat"' (once) without any sort of evaluation of the evidence related to it or recognition that the fault of something being exposed might lie with nVidia.
I guess this "balances" the prior recognition of the possibility of fault with nVidia further down the page. (the problem being, that recognition doesn't need to be "balanced").

A selection of words and phrases that are related by the text in question to the behavior of these entitities (other than nVidia): "strange things happening", "fails to register the true meaning","See where I'm going?", "convenient outlets", "misleading description", "Whistle Blowers". The irony of all of those except the last...which simply seems juvenile.

 

[Mariner]

Well, it looks to me as though NVidia have just decided to batten down the hatches and ride this storm out without any response.

You would have thought that their much-vaunted driver team could have found out the cause of the problems by now? It's been a week now since ExtremeTech and B3D first mentioned published any mention of the apparent driver problems on the web and we've had no response whatsoever from NVidia other than "we think it's a driver bug".

Normally in one week, we'd expect the leakage of at least two or three beta versions of Detonator drivers, wouldn't we?

 

[Doomtrooper]

Is the 44.03 drivers allowed to upload scores to the ORB, as if they can that is one way to get some response...no scoring should be allowed with these drivers.

 

[demalion]

Innuendo:
Well, we've had no response that has been stated to be directly from nVidia.
We know nVidia PR contacts websites and is willing to spell out the "party line" in great detail. With the apparent lack of sensibility or coherency (independent of nVidia's self interest) in some responses by websites, they seem to share some common characteristics with the prior known examples from nVidia in this regard.


Strange how far you can go with innuendo...I'd even dare say that since I can actually point to some responses and illustrate an actual lack of sensibility and coherency displayed, I've even provided more support for mine.

That's just my opinion, though, and I'm just a poster on a forum.

 

[galperi1]

what I don't understand is why Futuremark is NOT making a public statement. They are allowing their name to get drudged through mud by websites such as [H] after it was Nvidia's fault for inserting clipping planes in THEIR benchmark. If it were me, I would definately go on the offensive rather than stay quiet like they are doing. The more time that passes that nothing happens, the closer to other sites deciding 3dmark03 is a worthless benchmark because of Nvidia's dedication to proving it is easy not make equal comparisons between cards which is exactly what 3dmark is supposed to do just like any other standard benchmark

 

[AJ]

what I don't understand is why Futuremark is NOT making a public statement.

We're still doing research about the matter so for now we'll have to stick with the 'we may or may not comment this issue later on'.

Cheers,

AJ

 

[Slides]

You mean it's not clear as day that Nvidia cheated?

 

[jjayb]

I get minor differences through all tests on the NV30.

However, If I run this through the developer version then this clears one of the issues. The effect that is most notable in the first screenshots is cleared up.

Which effect? The proxycon issues or the Nature test issues? How much of a difference does it make on that test when the issues are gone?

 

[Evildeus]

...
By admitting those two new partners, FutureMark gets two very convenient outlets when they wish to issue slaps within the industry. This is something they have never had the ability to do. Nice huh? Convenient too, no? Certainly interesting. Assassins for hire?
...

Something at 3DChipset that reminds me of Kyle's rant.

Look at the answer of AJ:

'we may or may not comment this issue later on'