Is NVIDIA cheating to sell graphic cards?

[marcocom]

Originally posted by The Baron
The better question is do synthetic benchmarks matter.

it finally proves without a doubt that as a test of game performance, synthetic benchmarks are dead. And I think it's article time. ;)

indeed, you are right sir.

WHQL means alot more than Futuremarks' input to me. And then, how about that whole 'nvidia has no access to source code' fact there...nvidia cheating?? how about simply 'ATi knows how to please the benchmark precisely while nvidia is still guessing'

frankly, i just wish the benchmark didnt exist and that my fellow consumers stopped treating it like its gold. cuz its a joke. the idea that a benchmark exists that CERTAIN hardware mfrs have more access to than others certainly stinks more than anything regarding missing effects in playback of that benchmark by nvidia.

[Behemoth]

Originally posted by MuFu
It's not just about 3dm - they can do this for ANY timedemo, so people can get a inflated view of performance in actual games.

MuFu.
hmm ok, if they do it only on timedemo, thats bad, thats cheating, but as long as the actual games run great, i dont care too much about 3DMark.

[The Baron]

OK, I'm working on a way to randomly create a benchmark (essentially a timedemo) in UT2k3 that will still provide easily replicable results.

Will let you know how it's going once I know. ;)

[jjjayb]

Jeez you people are missing the point. If they do it on 3dmark2003 they can do it on a timedemo too. Most of the review sites run the same benchmarks. 3dmark2003, 2001, ut2003, splinter cell, serious sam etc... Guess what, they use the same time demos for these games too. ut2003 flyby sound familiar? If they do this for the time demos then they will get inflated scores. You'll buy a $500 card thinking it can get 60Fps in splintercell because that's what the benchmarks said, yet when you take the game home to PLAY it, it only plays at 40fps.

If they could make this hack work on a game that would be great. You'd have great fps while still thinking it looks good. Unfortunately it doesn't work that way.

Because they are doing this on a benchmark that makes it even worse. A benchmarks only purpose is to test a cards speed. They are inflating the speeds to sell more cards. You think you're going to run pixel shader 2.0 games great because they score great on p.s.2.0 benchmarks, but come to find out you spent $500 on a card that runs pixel shader 2.0 games like ass.

Kills me you people actually defend this type of action. I can understand having brand loyalty, but at some point you have to open your eyes.

[night]

Originally posted by Behemoth
hmm ok, if they do it only on timedemo, thats bad, thats cheating, but as long as the actual games run great, i dont care too much about 3DMark.

that is is more idiotic than NV40's post. i think... cant be far from it....
im gonna have to call that a tie actually

[Rogozhin]

well said jjj

maybe a similar analogy would be car makers using higher compression on all the cars they send out as new models to car magazines and reviewers then after the scores have been published lowering the compression (saving the stress on the engine) and selling them to the public.

This could easily happen if nvidia is allowed to do this sort of thing.

rogo

[ChrisRay]

I don't think anyone said benchmarks are dead, But synthetic benchmarks are slowly dying out of the PC industry.


People have tried to stay away from them. Expecially since most games have benchmarks now

[Behemoth]

Originally posted by jjjayb
Jeez you people are missing the point. If they do it on 3dmark2003 they can do it on a timedemo too. Most of the review sites run the same benchmarks. 3dmark2003, 2001, ut2003, splinter cell, serious sam etc... Guess what, they use the same time demos for these games too. ut2003 flyby sound familiar? If they do this for the time demos then they will get inflated scores. You'll buy a $500 card thinking it can get 60Fps in splintercell because that's what the benchmarks said, yet when you take the game home to PLAY it, it only plays at 40fps.

If they could make this hack work on a game that would be great. You'd have great fps while still thinking it looks good. Unfortunately it doesn't work that way.

Because they are doing this on a benchmark that makes it even worse. A benchmarks only purpose is to test a cards speed. They are inflating the speeds to sell more cards. You think you're going to run pixel shader 2.0 games great because they score great on p.s.2.0 benchmarks, but come to find out you spent $500 on a card that runs pixel shader 2.0 games like ass.

Kills me you people actually defend this type of action. I can understand having brand loyalty, but at some point you have to open your eyes.
i believe ati is well capable of doing this kind of benchmark hack too, just because it can doesnt mean it will do it.
so does 3Dmark difference indicate splinter cell performance difference? is there any reason why nvidia still hasnt hacked the AA/AF performance to beat 9700/9800pro?
3DMark is not supposed to mean absolute performance of every single game. it is far from an absolute performance indicator of every single game.

[Nv40]

Originally posted by marcocom
indeed, you are right sir.

WHQL means alot more than Futuremarks' input to me. And then, how about that whole 'nvidia has no access to source code' fact there...nvidia cheating?? how about simply 'ATi knows how to please the benchmark precisely while nvidia is still guessing'

frankly, i just wish the benchmark didnt exist and that my fellow consumers stopped treating it like its gold. cuz its a joke. the idea that a benchmark exists that CERTAIN hardware mfrs have more access to than others certainly stinks more than anything regarding missing effects in playback of that benchmark by nvidia.


indeed..

ATi have plenty of Access to 3dmark2003 source code ..
while Nvidia not.. so they can "optimize" the way they want
to get better scores , just for 3dmark2003..
i dont see ,the fairness here... at all .

so thats why i dont care too much about numbers in
Syntetic benchmarks like 3dmarks /SHadersmarks/RIghtmarks
and others.. They may be good as graphic Demo , but in no way
represent my future performance in 3d games ,if my video
card company is not allowed to work with them ,the way it happens
in the real world with games.. ;)

the benchamrk can be "legal" and at the same time manipulated
their programming to benefits more ,one IHV video card
arquitecture . (cough) PS1.4 video cards..

if Nvidia Nv30 were not delayed by 6months ,the
"standar" now in DIrectx9 will be other.

for me the real BEnchmarks are games.,
and here is where ATI now have a new trouble . :)

[Behemoth]

Originally posted by night
that is is more idiotic than NV40's post. i think... cant be far from it....
im gonna have to call that a tie actually
your post looks purely fanatical to me as well

[Richteralan]

I really have to congratulation to ATI's PA.......
They done a great job to make people thinking Nvidia is cheating.....hahaha.........

[gokickrocks]

:rolleyes:

i have been spectating for awhile now and feel that i need to chime in...

has anyone noticed that once the nv35 got caught cheating, some people on this forum act like it never happend and are accusing ati of doing a hack of the same degree? then there are also those that say it is just nvidia's way of trying to show that they can optimize, yet when they were confronted with the issue, they call it a "bug" and are investigating the issue...if they were trying to prove a point, wouldnt they outright proclaim it?...

those same people that were gloating about the nv35's score in 3dmark are now condemning futuremark...you cant have it both ways

then there are some that believe that ati does poorly in games...it seems fine to me and i run it at 6xAA 16xAF 1280x1024 and i never see the screen come to a crawl...but my card must be 1 in a million, cause those people that dont own it, surely must know better

[mreman4k]

Originally posted by Nv40
indeed..

ATi have plenty of Access to 3dmark2003 source code ..
while Nvidia not.. so they can "optimize" the way the want
to get better scores ,just for 3dmark2003..
i dont see ,the fairness here... at all .

so that why i dont care too much about numbers in
Syntetic benchmarks like 3dmarks /SHaders marks/RIghtmarks
and others..

they can be "legal" and at the same time manipulated
their programming to benefits more ,one IHV video card
arquitecture . (cough) PS1.4 video cards..

if Nvidia Nv30 were not delayed by 6months ,the
"standar" now in DIrectx9 will be other.

for me the real BEnchmarks are games.,
and here is where ATI now have trouble . :)

nVIDIA specifically chose not to be able to access the source code so thats in their hands. This is an outright cheat and is not morally right in giving the average consumer clouded conclusions of what the best video card is. But who is morally right nowadays? That is why I choose not to give my money to nVIDIA...all goes back to the GF4 MX thing...everyone cheats though, but I choose to make an example out of nVIDIA...rofl :D

[Behemoth]

3DMark itself is bad enough for average gamer who doesnt understand how good the actual game card is capable of in real world performance due to things like special path and special support, due to its a vague indication of actual real game performance.

[Moose]

Originally posted by Nv40
indeed..

ATi have plenty of Access to 3dmark2003 source code ..
while Nvidia not.. so they can "optimize" the way they want
to get better scores , just for 3dmark2003..
i dont see ,the fairness here... at all .



huh???

ATI can optimize all they want yet it is Nvidia who have optimized the hell out of this benchmark to the point of creating a clip plane in every frame of the benchmark.

That must have been an enormous ammount of work. All for the sole purpose of cheating the benchmark and deceiving consumers. There is no way to sugar coat it at all.

[StealthHawk]

Let me chime in here.

Doing this in a synthetic benchmark is totally unfair. It is cheating, as it artificially inflates scores in comparison to other cards, which render everything.

Now, is this cheating if done in GAME benchmarks like timedemos? Maybe yes, maybe no. I think if NV3x cards do the same thing when you're actually playing a game, then no, it's not cheating. Because it is representative of real world performance.

Looks like nvidia got something out of 3dfx after all. 3dfx's uber 1337 software HSR routines :p

[ChrisW]

Let's not forget that nVidia was a beta member of Futuremark until December. They chose not to continue their membership because they would not optimize their code specifically for nVidia cards. As far as source code goes...do you think they destroyed all the information they had about 3DMark03? It's not like they didn't know exactly how it worked in December.

[PreservedSwine]

I wonder if Nvidia will "recall" these drivers? Not likely, but damn, that is one BIG error. I can already see all the screenshots floating aroufn the web as soon as you go "off the rail" so to speak.

[Lucien1964]

MUHAHAHAHAHAHAHAHAHA!!!!!! I remember the day when Nvidia swore by 3dmark.I'm sure the 5900 is a kick @ss card.Why cheat???It will just hurt them more getting caught cheating.I dunno peeps but trusting Nvidia again will be a tough task.

[jjjayb]

so does 3Dmark difference indicate splinter cell performance difference?

No. But if they are running this hack on 3dmark03 then they could just as well run it on the splinter cell timedemos. Then it could be that the Splinter Cell timedemo benchmarks aren't indicitave of real Splinter Cell performance when you are actually playing.

[bkswaney]

My 3DM03 scores are the same with the 43.51 WHQL and Det 44.03's.
They might very well have a bug but it's not messing up it's score.
Now my games and 3DM01 went way up. :)

I think ET is just trying to stir up $h!t.

Just my 2 cents.
Also I've not see any problems in 03 myself.

[bkswaney]

Originally posted by Nv40
yes , but we have already sites and Users Claiming that
Nvidia latest DEtonatorsFX drivers Fixed
3dmark2003 issues...

so until you can proof me that those "slight" errors ,
replicates in EVERYONE!!! using the DEtFx ,
then i see the Extremetech report,
Baseless on non FUndamental FActs..

if those "slight" errors does not happens in everyone
using the DetonatorsFx then this clearly means is not
in his drivers the Issue.

I agree... I've run it 6 times and now I see nothing wrong.
My IQ is BETTER than with the 43.51's. Speed about the same.
Maybe 100 points higher.

Plus these are WHQL cert. They must meet MS's eye. :)

[MrNasty]

I must say that what nvidia has done is downright appaling.


Effectivly trying to pull the wool over the entire gamer community in the hope to make a quick buck.

They flipping hand crafted the PRE RENDERED culling routines to remove so much information.

What nvidia has done is invalidated the becnhmark, heck they might as well of included a prerendered mpeg of all the scenes and just stuck that in and culled the rest.

Ultimately dispicable and really is proof of my distrust of nvidia and in general corporations at large, Im sure ATi is no clean slate either.

Hypoicarcy comes to mind "we don't care about 3dmark because they wont include special rendering routines to make the benchmark look different and run different on our card"

"so we will sneak them into our drivers"

[Phyre]

Thanks for trying the different drivers, bkswaney. I was wondering when someone was going to do that.

Phyre

[PreservedSwine]

Are there any benches of the FX5900 using both driver revisions in 3dm03?:confused:

I think we'll find out eventually, this *might* be NV35 specific.