Nvidia is up to their old tricks

[Technoid]

According to extremetech nvidia is using cheat drivers to make the GF FX 5900 seem faster in 3dmark2003
Read it

[Jammrock]

What if I don't want to read it?

Knowing nVidia they paid off Mad Onion to put that proprietary nVidia crap in 3dmark 2003 that's in games like UT2k3.

"Driver bug" ... my big fat hairy white arse it's a driver bug. They thought they could get away with it.

Jammrock

[agallag]

Kyle at HardOCP has a different opinion about the whole mess. It'll be interesting to find out what's really going on...

[cjolley]

Oh come on guys!
This happens to me all the time.
I make a mistake in my code and the app runs better!
That's how good programming works!
chuck

[[GDI]Raptor]

I couldn't care less if nVidia have some "rendering problems" in 3DMark. It's not a real life application.

I think this is what we call a driver bug! And the fanATIc's are going completely mad over it.... Just because NV35 beats Radeon 9800 Pro in that menchmark......

I have to admit that I'm glad Matrox isn't a part of this "look I get one fps more than you" game....

[bsdgeek]

I dont see how this matters AT ALL. So nVidia doesn't render the pixels you can't see? WTF 1 W4/\/+ /\/\1551/\/g p1x31z! The only way to see this, so called, cheating is to use the Beta 3dMark, which you can't have.

[Technoid]

okay, one small clarification:
I use a MSI GF4 4200 !!

The only thing I could admit to being a "Fanboy" to is Litestep

[Marshmallowman]

hmm...quack was with unoffical beta drivers, but it was dodgy.

This NV one is with official drivers, there seems to be hardcoded static clipping planes implemented by NV, eg the card does no work finding some of the polygons to cull. Effectively invaliding this test for NV hardware, because in a game you could not have they could not be static...it is very dodgy, defintely inflating their score.

NV has been associated with so many questioinable things lately, I was actually considering an new NV card for myself(mainly due to linux). But there whole attitidue just screams panic and BS.

They want to have the fastest card no matter what....

When they originally released there paper nv30 it was "apparently" twice as fast as the best ATI had...NO not even the nv35 is that fast. WAY to much BS for me, even though nv30 was the card of the year for 2002

[Drizzt]

Originally posted by bsdgeek
I dont see how this matters AT ALL. So nVidia doesn't render the pixels you can't see? WTF 1 W4/\/+ /\/\1551/\/g p1x31z! The only way to see this, so called, cheating is to use the Beta 3dMark, which you can't have.

Totally wrong.
nVidia doesn't render the pixels that they already know are invisible. This is cheating.

[bsdgeek]

Drizzt, that's what I had meant if not what I said. Only members of the 3DMark beta can fly around and see the areas not being rendered. Since I'm not a 3DMark beta member, I can't fly around to see those. So does it matter?
For all I care about this, ATi & Matrox should go ahead and do it.

What about Tile Based Renderers, they remove the pixels you can't see, do you think that is cheating?

[Drizzt]

Originally posted by bsdgeek
Drizzt, that's what I had meant if not what I said. Only members of the 3DMark beta can fly around and see the areas not being rendered. Since I'm not a 3DMark beta member, I can't fly around to see those. So does it matter?

Ok, so I think you would like to play an FPS and be unable to move the head around because nVidia drivers doesn't allow you to do it, right?




Edit: I have problems xplaining this with my poor english...can someone else explain the matter to bsdgeek?

[dbdg]

Some idiots buy a card based on 3D Mark scores and assume that this reflects game performance, with cheating like this there is no link between 3D Mark performance and game performance as there would be with ATI cards.

Then again the idiots that buy cards based on 3D Mark scores deserve to get this kind of treatment.

[treilanin]

You know what, I am hardly surprised by the posturing NVidia is taking by doing this. I was hardly a NVidia fan before and even though I am an NVidia user at this point (MSI GF4 4200), I am still less then impressed with their drivers (Linux Driver excepted). Why do I say this? NVidia has always made gradiose claims that a driver release will suddenly be able to give 25% over the previous generation..... The fanboys just lap that statement up, but really have we ever stopped to think about the statement in total?

I have no urge to ever buy another NVidia card again. Install new drivers and get worse graphic performance... not exactly something that I condone or want. To me the benchmarks given on most sites are totally useless and irrelevent since they only focus on most FPS or most Render points or highest 3DMARK score(typical fanboy review sites like [H]OCP do this to a letter). What I want to see is more emphasis on graphic quality, average/mean/median FPS and the deviation of those values. I am pretty sure you find the video card landscape to be much different that.


To BSDGEEK:

What NVidia is doing is entirely cheating, they have gone in and analyze a static demo and optimized the drivers to occlude pixels. Now you may say that is fine and all... more power to NVidia. Well how would you feel if you put down $600 for a brand spanking new 5900 Ultra because it had 1500 more 3DMarks over the $500 ATI card only to find that the ATI does everything the same in every game? You would not be happy at all.

As for claiming that tile based renderers are cheating... that must be the most knee jerk statement to make to justify the crap that NVidia pulls. A tile based render would never be considered a cheat because the benefits that it gives to 3DMark are equally applied to all games that run on it.

Perhaps you just play games differently then the majority of us? Maybe you play a version of games where you can't look around, you can't do anything, you only move when the machine moves you.... in that case, yes the NVidia driver hacks are just perfect for you I on the other hand like to be able to move on my own in the games I play :P

[TdB]

What about Tile Based Renderers, they remove the pixels you can't see, do you think that is cheating?

well, tile based rendering actually works for (allmost) all games, Nvidia´s method only works for very specific scenes seen from very specific angles, and only in that specific benchmark. they pretty much placed additional clipping planes by hand, for that specific benchmark run(if I understand correctly).

these kind of optimizations won´t work for real games. let me try to explain why:

lets say they did the same optimizations for a first person shooter, and lets say that benchmarkers uses a timedemo for that game, this timedemo takes place in room A.
the optimization is a crude hack the says: "don´t draw anything, but room A, ever!"(its a bit more complicated than that, but thats the idea). this will work well for that particular timedemo, but if you actually play the game, you might actually want to go to room B, which won´t be rendered as long as the drivers uses that optimization, thus making the game unplayable.

tilebased rendering works quite differently, (for each frame, before it actually draws anything) it examines what polygons are hidden behind other polygons, and then based on that examination it decides which polygons to render and draw. This optimization will work on allmost all games(there are a few exceptions) and thus benefit the end user.

the problem with what nvidia is doing is that, they trick their potential custumers into believing that their cards are faster than they actaully are. Another problem is that these hacks are timeconsuming to create, this time could have been spend on REAL optimizations and bugfixes that would actually benefit their costumers.
Spending precious time on cheating in benchmarks instead of actually improve their drivers is (IMO) very bad service, and downright disrespectfull to their costumers.

BTW, claiming a 30% performance boost in drivers is the same as admitting, that their previous drivers was incredible inefficient and badly coded. and since I don´t think their drivers are that bad, then it is quite obvious that the performance boost is somewhat shady.

[bsdgeek]

Hey look, I never said anything about the way I play games. I said that I don't think it matters what nVidia is doing in 3DMark. I had not of anyone thinking 3DMark is worth a damn anyway. nVidia has done nothing, so far, in real games to do anything. Now if they were going to plant some static clip planes in a real game, I might be angry.

I was probably wrong about mentioning TBRs, but I was trying to find some other example.