All fingers point to Nvidia doing some cheating with 3DMark2003.Check out the link for a good read 4-5 pages long. Your Cheatin Heart 
Nvidia Busted-Cheating With Their New FX Drivers
- Thread starter LastRide
- Start date
meh, i personally dont care if one graphics company or another 'cheats' their video cards or drivers just to get a better score on 3dmark. its just a general video card benchmarking program. all that matters is how the cards perform in real games
well.. thats my opinion
well.. thats my opinion
the problem is that if nvidia can optimize drivers for games within api specs then comarisions with other vendors cards are MOOT .
Hopefully this will make reviewers and consumers more skeptical of "optimizations".
I've noticed many reviewers that have stange discrepencies in their benchmarks and it is a relief to find ONE site at least that IS looking into these things.
I also believe that had ati done this people and sites would be harping about for a long time, but as it stands it seems that nvidia can rely on it's market share and sheer corporate size to pull itself through nasty mud like this.
I'd say that this sheds alot of light on the nvidia's integrity in the poweruser community.
I don't see how this could be a "bug" as they claim since 3dmark2003 has been out for a few months and this is nvidia's WHQL driver.
There is a little poll going on at nvnews.net about this and if you fellas would like go check it out.
Rogo
Hopefully this will make reviewers and consumers more skeptical of "optimizations".
I've noticed many reviewers that have stange discrepencies in their benchmarks and it is a relief to find ONE site at least that IS looking into these things.
I also believe that had ati done this people and sites would be harping about for a long time, but as it stands it seems that nvidia can rely on it's market share and sheer corporate size to pull itself through nasty mud like this.
I'd say that this sheds alot of light on the nvidia's integrity in the poweruser community.
I don't see how this could be a "bug" as they claim since 3dmark2003 has been out for a few months and this is nvidia's WHQL driver.
There is a little poll going on at nvnews.net about this and if you fellas would like go check it out.
Rogo
Well, I guess (if that's true) that the nvidiots won't be able to point a finger at that fanATIcs anymore now that both companies have cheated at one time or another in some benchmark or application.
This is very interesting. Good scoop LastRide.
3DgameGage consists of 8 game tests ?
Comanche 4 (Direct3D)
Dungeon Siege (Direct3D)
IL-2 Sturmovik (OpenGL)
Jedi Knight II (OpenGL)
NASCAR 2002 (Direct3D)
NHL 2002 (Direct3D)
Serious Sam SE (OpenGL)
Unreal Tournament 2003 (Direct3D)
The 9800 was faster in 5 of these 8 game tests and its overall score was faster.
extremetech ?
Exactly, and in the game tests that Nvidia very likely didn?t optimize for, the 5900U got beat by the 9800.all that matters is how the cards perform in real games
3DgameGage consists of 8 game tests ?
Comanche 4 (Direct3D)
Dungeon Siege (Direct3D)
IL-2 Sturmovik (OpenGL)
Jedi Knight II (OpenGL)
NASCAR 2002 (Direct3D)
NHL 2002 (Direct3D)
Serious Sam SE (OpenGL)
Unreal Tournament 2003 (Direct3D)
The 9800 was faster in 5 of these 8 game tests and its overall score was faster.
extremetech ?
Schadenfroh
Elite Member
- Mar 8, 2003
- 38,416
- 4
- 0
Originally posted by: Vegetto
meh, i personally dont care if one graphics company or another 'cheats' their video cards or drivers just to get a better score on 3dmark. its just a general video card benchmarking program. all that matters is how the cards perform in real games
well.. thats my opinion
I completley agree
ATI cheated in quake3
Nvidia cheats in 3dmark 2003
which one is worse when you make a video card purchase to play games?
I know ATI cheated once,but as far as I know this is the second time Nvidia cheated that I remember.Nvidia did it a while back.I remember talking to my cousin about it.
how deep does the rabit hole go???
I'd bet money that the doom3 beta is fubared like this-as well as the ut2003 bot match benchies and sc bench.
Anytime you use a track camera doing flyby or fps orientation then this comes into play.
So you wouldn't care if when you bought the card and ran a game "the way it way meant to be played" it performed much worse than the reviews stated?
I also think it's wrong for nvidia to send out a whql driver to reviews sites where time and money is Wasted because of a direct manipulation of a benchmark.
Rogo
I'd bet money that the doom3 beta is fubared like this-as well as the ut2003 bot match benchies and sc bench.
Anytime you use a track camera doing flyby or fps orientation then this comes into play.
So you wouldn't care if when you bought the card and ran a game "the way it way meant to be played
" it performed much worse than the reviews stated? I also think it's wrong for nvidia to send out a whql driver to reviews sites where time and money is Wasted because of a direct manipulation of a benchmark.
Rogo
Shamrock
Golden Member
- Oct 11, 1999
- 1,441
- 567
- 136
Now the dilemma...Did NVidia cheat in 3dMark 2k3, or are they showing us "flaws" in the program? I would see it as it's a flaw in 3dmark...because AMD and NVidia both have been outspoken about 3dMark.
NVidia isnt cheating at anything, they are exposing a flaw in the program. At least NVidia didnt "cheat" in a game result, and only exposed a flaw in a SYNTHETIC benchmark :\
NVidia isnt cheating at anything, they are exposing a flaw in the program. At least NVidia didnt "cheat" in a game result, and only exposed a flaw in a SYNTHETIC benchmark :\
Keysplayr
Elite Member
- Jan 16, 2003
- 21,211
- 50
- 91
If any of you owned a graphics company like nvidia or ATI, and you had to "cheat" to win in a single benchmark program (3dmark2003),
and your reputation is tarnished from a bad release (NV30), you would be fooling yourself if you say that you would not "optimize (cheat)" to win.
I know I would. But guys, its only 3dmark2003 and its just a rumor anyway. Isn't it?
Who here thinks the NV35 is a piece of crap? Show of hands please.... If you raised your hand, you are out of your mind.
Who here thinks the NV30 is a piece of crap? Yeah sure, its loud and hot and takes up 2 slots in your case, but it kicked arse as well.
Now even more so with AA and AF fixed. And look at the 5600 ultra. I couldn't believe it is beating out the 9500 pro and 9600 pro now.
We are talking kick-arse technology from both sides. Enjoy it. Dont nit-pick.
Keys
and your reputation is tarnished from a bad release (NV30), you would be fooling yourself if you say that you would not "optimize (cheat)" to win.
I know I would. But guys, its only 3dmark2003 and its just a rumor anyway. Isn't it?
Who here thinks the NV35 is a piece of crap? Show of hands please.... If you raised your hand, you are out of your mind.
Who here thinks the NV30 is a piece of crap? Yeah sure, its loud and hot and takes up 2 slots in your case, but it kicked arse as well.
Now even more so with AA and AF fixed. And look at the 5600 ultra. I couldn't believe it is beating out the 9500 pro and 9600 pro now.
We are talking kick-arse technology from both sides. Enjoy it. Dont nit-pick.
Keys
1. What seems like a long time ago, ATi "cheated" on Quake 3, but later fixed their "error," and we ended up with perfect IQ at "cheat" speeds--so their explanation of a driver bug seems half-credible.
2. FX does incredibly poorly in 3DM03 compared to 9700P. nV releases new Dets that increase performance drastically and tells sites to use them to benchmark with. Sites soon realize that nVidia cheated on 3DM03 with their obviously incorrectly rendered sky, whch people took to mean that they were using INT rather than FP precision, or at least FP16 rather than at least FP24 (which would be FP32, in their case). nVidia "fixed" the sky problem with the 43.00 and 43.03 drivers, but scores dropped down to pre-cheat levels. Finally nV releases 43.51 drivers: IQ is fixed and I believe scores remained as high as with the cheat.
(I'll ignore the fact that nV also pressed review sites to compare their "Quality" AF to ATi's "Quality". Note that at the time, nV's "Quality" was worse than their current 44.03 setting of "Performance.")
3. Now this. We'll see if they can produce a driver which ET and other 3DM03 Beta members can verify doesn't introduce unacceptable culling.
It's funny that nVidia themselves said there were lots of ways to cheat 3DM03, and that's one of the reasons why they didn't trust it as a benchmark. I guess they showed us that they were telling the truth, right? (Just like when they said NV30 was eight pixels per clock.) This also raises the spectre that they might have done the same thing with their prepared Doom 3 demo--though since they agreed to let id create a new one, since id wasn't happy with nV's, I don't think nV would've had the time to implement some extra-game culling for this D3 benchmark. They've obviously been concentrating on it quite a bit either way, and they could've bought a 9800P off the shelf to test against their NV35, so they must've been confident they would win (significantly) no matter what.
nVidia puts out great hardware. It's these lapses of ethical judgement that gets to me, though obviously more so than others. It'd be nice for everyone involved to discover this was simply a mistake, but nV's recent history seems to indicate otherwise.
I'm sure Anand will put this in his Det FX article, though, so at least we know we're getting some good reading material out of this.
2. FX does incredibly poorly in 3DM03 compared to 9700P. nV releases new Dets that increase performance drastically and tells sites to use them to benchmark with. Sites soon realize that nVidia cheated on 3DM03 with their obviously incorrectly rendered sky, whch people took to mean that they were using INT rather than FP precision, or at least FP16 rather than at least FP24 (which would be FP32, in their case). nVidia "fixed" the sky problem with the 43.00 and 43.03 drivers, but scores dropped down to pre-cheat levels. Finally nV releases 43.51 drivers: IQ is fixed and I believe scores remained as high as with the cheat.
(I'll ignore the fact that nV also pressed review sites to compare their "Quality" AF to ATi's "Quality". Note that at the time, nV's "Quality" was worse than their current 44.03 setting of "Performance.")
3. Now this. We'll see if they can produce a driver which ET and other 3DM03 Beta members can verify doesn't introduce unacceptable culling.
It's funny that nVidia themselves said there were lots of ways to cheat 3DM03, and that's one of the reasons why they didn't trust it as a benchmark. I guess they showed us that they were telling the truth, right? (Just like when they said NV30 was eight pixels per clock.) This also raises the spectre that they might have done the same thing with their prepared Doom 3 demo--though since they agreed to let id create a new one, since id wasn't happy with nV's, I don't think nV would've had the time to implement some extra-game culling for this D3 benchmark. They've obviously been concentrating on it quite a bit either way, and they could've bought a 9800P off the shelf to test against their NV35, so they must've been confident they would win (significantly) no matter what.
nVidia puts out great hardware. It's these lapses of ethical judgement that gets to me, though obviously more so than others. It'd be nice for everyone involved to discover this was simply a mistake, but nV's recent history seems to indicate otherwise.
I'm sure Anand will put this in his Det FX article, though, so at least we know we're getting some good reading material out of this.
I don't know if I can even care about anything on that site, I can't believe how WORTHLESS their 5900 review is.
Way to pick the ONLY benchmarks you would NEVER run these cards at
Good Lord. Someone needs to tell these guys NO FPS PLAYER IN THE WORLD WOULD RUN UT2003 AT 20fps?!?!? What are we supposed to get from this? That the cards are pretty evenly matched if you run them at bizarre slide show settings that would induce epilepsy in most people?
Way to pick the ONLY benchmarks you would NEVER run these cards at
Good Lord. Someone needs to tell these guys NO FPS PLAYER IN THE WORLD WOULD RUN UT2003 AT 20fps?!?!? What are we supposed to get from this? That the cards are pretty evenly matched if you run them at bizarre slide show settings that would induce epilepsy in most people?
You claim that's this doesn't matter since it isn't visibly noticable in games...but that's not the point.
'This is a BENCHMARK, and furthermore it's the specific benchmark that nVidia shunned and called meaningless and said they would never optimize drivers for...and that article is VERY CLEAR EVIDENCE that they have indeed been playing very fast and loose with their driver optimizations to cheat their way to a better score.
Not that big of a deal? Well, the two big stories yesterday that I recall reading were the FX slaughtering the 9800 in Doom3 and 3dm2k3. D3 I already know the reasons behind, and while it wasn't cheating it was more a business maneuver than a fair shoot-out in card performance...but THIS one is flat-out cheating.'
I agree with the extremetech gamegauge and that these synthetic benchmarks are pretty worthless but this whole episode with nv tweaking drivers for more speed is simply begging the question-where does it stop?
rogo
'This is a BENCHMARK, and furthermore it's the specific benchmark that nVidia shunned and called meaningless and said they would never optimize drivers for...and that article is VERY CLEAR EVIDENCE that they have indeed been playing very fast and loose with their driver optimizations to cheat their way to a better score.
Not that big of a deal? Well, the two big stories yesterday that I recall reading were the FX slaughtering the 9800 in Doom3 and 3dm2k3. D3 I already know the reasons behind, and while it wasn't cheating it was more a business maneuver than a fair shoot-out in card performance...but THIS one is flat-out cheating.'
I agree with the extremetech gamegauge and that these synthetic benchmarks are pretty worthless but this whole episode with nv tweaking drivers for more speed is simply begging the question-where does it stop?
rogo
Rollo, these were all previews. Reviewers had limited time with the cards in which to release a p/review on the day the NDA expired. Many other sites, like Anand, said they would test at only 4xAA and 8xAF to stress these $500, 23+GB/s cards--not to mention high res and high AA would also emphasize their 256MB's of memory. ET made their intentions pretty clear:
I don't see you complaining too loudly that the low-end cards got only 12 and 20fps at Doom 3 at Anand's lowest setting of 1024x768 Medium Quality....
Rollo
"Good Lord. Someone needs to tell these guys NO FPS PLAYER IN THE WORLD WOULD RUN UT2003 AT 20fps?!?!? What are we supposed to get from this? That the cards are pretty evenly matched if you run them at bizarre slide show settings that would induce epilepsy in most people."
It doesn't on the 9800 pro, it ran at 27fps @1600 and 38fps @1600. Would you rather have them test it at lower resolutions (I thought you liked running hi-res since you don't like AA or AF that would be the only other option to increase IQ) where the cpu becomes more of a limiting factor? As they mentioned the cpu could possibly be holding all three video cards back by a 50% performance margin.
It's one of the better review sites around since they use many different games, a few synthetic tests, and their "gamegauge."
Also
You used a logical fallacy again- I "don't know if I can even care about anything on that site, I can't believe how WORTHLESS their 5900 review is."
Basing your conclusion on ONE review is the fallacy called "unlogichno rollo"
Rogo
"Good Lord. Someone needs to tell these guys NO FPS PLAYER IN THE WORLD WOULD RUN UT2003 AT 20fps?!?!? What are we supposed to get from this? That the cards are pretty evenly matched if you run them at bizarre slide show settings that would induce epilepsy in most people."
It doesn't on the 9800 pro, it ran at 27fps @1600 and 38fps @1600. Would you rather have them test it at lower resolutions (I thought you liked running hi-res since you don't like AA or AF that would be the only other option to increase IQ) where the cpu becomes more of a limiting factor? As they mentioned the cpu could possibly be holding all three video cards back by a 50% performance margin.
It's one of the better review sites around since they use many different games, a few synthetic tests, and their "gamegauge."
Also
You used a logical fallacy again- I "don't know if I can even care about anything on that site, I can't believe how WORTHLESS their 5900 review is."
Basing your conclusion on ONE review is the fallacy called "unlogichno rollo"
Rogo
Keysplayr
Elite Member
- Jan 16, 2003
- 21,211
- 50
- 91
Yes I would Pete. And if you had millions of dollars at stake, I'm sure you would find your streetsmarts soon enough.
Who knows how the CEO of Nvidia feels. I will never know. Does he sleep well at night? Don't care. Does he make great graphics
boards and chipsets? Do care.
Dont get all holy'r than thou' on us. Your not. Especially the ones that pretend to be.
Keys
Who knows how the CEO of Nvidia feels. I will never know. Does he sleep well at night? Don't care. Does he make great graphics
boards and chipsets? Do care.
Dont get all holy'r than thou' on us. Your not. Especially the ones that pretend to be.
Keys
bad website, good website, bad benchmarks, good benchmarks, 3dmark bad, 3dmark good...
what does it matter? If they cheated, I don't care how good their product is, its despicable.:disgust:
But I'm holding out on judgement just because I think the claims have been made are way too bold and
much too fast.
what does it matter? If they cheated, I don't care how good their product is, its despicable.:disgust:
But I'm holding out on judgement just because I think the claims have been made are way too bold and
much too fast.
This was said by david kirk a while ago (tongue in cheek!)
"We at NVIDIA don?t make it a practice to optimize our pipeline for specific benchmarks - we want to provide high quality and high performance on a wide variety of useful and entertaining applications, rather than just getting a good score. Ask yourself (or, better yet, ask ATI) why Radeon 8500 performs well on this one test, and poorly on many other 3DMark2001 tests."
Sure
rogo
"We at NVIDIA don?t make it a practice to optimize our pipeline for specific benchmarks - we want to provide high quality and high performance on a wide variety of useful and entertaining applications, rather than just getting a good score. Ask yourself (or, better yet, ask ATI) why Radeon 8500 performs well on this one test, and poorly on many other 3DMark2001 tests."
Sure
rogo
Shamrock
Golden Member
- Oct 11, 1999
- 1,441
- 567
- 136
NVidia isnt cheating, they simply exposed a flaw in 3dMark2k3. They knew tech sites would catch on.
and who's to say the"gamegauge" isnt ATI optimized? hmm... Extremetech seems biased towards ATI on the FIRST page
before you start calling me a fanboy or an NVidiot, I have owned every major video card, EXCEPT Matrox, I've even owned a Rendition based Diamond V2200....so I am completely neutral.
and who's to say the"gamegauge" isnt ATI optimized? hmm... Extremetech seems biased towards ATI on the FIRST page
before you start calling me a fanboy or an NVidiot, I have owned every major video card, EXCEPT Matrox, I've even owned a Rendition based Diamond V2200....so I am completely neutral.
No.
All the benchmarks including game demo bench are susceptible to cheats and hacks.
Then the whole benchmarks are useless? You know the answer.
And the synthetic benchmarks are good for its own accord.
They let us compare the systems on similar level and standardized situation.
No offense. But your comment sounds pointless.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 20K
-
-
-
Facebook
Twitter