Nvidia Busted-Cheating With Their New FX Drivers

[Blastman]

Once again I did read the article. What does it prove? It proves what Nvidia said a while ago that 3DMark means nothing and drivers can be optimized for it. 3dMark is not a game, its a scripted demo that goes over and over. So who cares!!!!!!

If they can "cheat" their drivers into giving more fps and I see no image loss, then let them "cheat" all they want. The same is for ATI, but as we all know, ?

The problem is that the cheat would only inflate the benchmark score, and the actual game would play slower than what the benchmarks indicate.

Likewise, if the UT3 bot match score was manipulated then the actual performance of your card in UT3 will be lower than what the benches indicate. This is not just about 3Dmark. This is about deceiving customers about the actual performance of their graphics cards. There is a lot at stake here $ wise too. One can only command top dollar for a product in the market place if it is perceived as a good product. The benchmarks have a lot to do with this.

Hardware sites are going to need to script their own benchmarks and perhaps use more games in the overall assessment.

 

[WicKeD]

This sheds some light on the situtation. I suppose I will have to suspend judgement until we see some more info.
rogo

I do agree with you on your comment. (A first today )

I do have to add that, the line needs to be drawn on what is optimizing and what is cheating. My opinion is if the scenes render faster and the IQ is good then it's an optimization Even if you have some tools that can show things that the normal user cannot see, such as rendering from different angles that show reduced IQ, to me its still not cheating. Cheating is clipping parts of the image that the average user can see to inflate fps, etc.

This is just my opinion any many will differ.

 

[WicKeD]

Originally posted by: Blastman

Once again I did read the article. What does it prove? It proves what Nvidia said a while ago that 3DMark means nothing and drivers can be optimized for it. 3dMark is not a game, its a scripted demo that goes over and over. So who cares!!!!!!

If they can "cheat" their drivers into giving more fps and I see no image loss, then let them "cheat" all they want. The same is for ATI, but as we all know, ?

The problem is that the cheat would only inflate the benchmark score, and the actual game would play slower than what the benchmarks indicate.

Likewise, if the UT3 bot match score was manipulated then the actual performance of your card in UT3 will be lower than what the benches indicate. This is not just about 3Dmark. This is about deceiving customers about the actual performance of their graphics cards. There is a lot at stake here $ wise too. One can only command top dollar for a product in the market place if it is perceived as a good product. The benchmarks have a lot to do with this.

Hardware sites are going to need to script their own benchmarks and perhaps use more games in the overall assessment.

True you are, but this is a norm in the industry. You really think the Jensen amps put out 1000 watts? No it is an inflated number. It puts out 1000 watts with a THD of 90%.

It does not matter how you get there, it's the numbers at the end that count. <--- Marketing Mind (not my mind)

 

[nutxo]

Originally posted by: WicKeD

Originally posted by: Blastman

Once again I did read the article. What does it prove? It proves what Nvidia said a while ago that 3DMark means nothing and drivers can be optimized for it. 3dMark is not a game, its a scripted demo that goes over and over. So who cares!!!!!!

If they can "cheat" their drivers into giving more fps and I see no image loss, then let them "cheat" all they want. The same is for ATI, but as we all know, ?

The problem is that the cheat would only inflate the benchmark score, and the actual game would play slower than what the benchmarks indicate.

Likewise, if the UT3 bot match score was manipulated then the actual performance of your card in UT3 will be lower than what the benches indicate. This is not just about 3Dmark. This is about deceiving customers about the actual performance of their graphics cards. There is a lot at stake here $ wise too. One can only command top dollar for a product in the market place if it is perceived as a good product. The benchmarks have a lot to do with this.

Hardware sites are going to need to script their own benchmarks and perhaps use more games in the overall assessment.

True you are, but this is a norm in the industry. You really think the Jensen amps put out 1000 watts? No it is an inflated number. It puts out 1000 watts with a THD of 90%.

It does not matter how you get there, it's the numbers at the end that count. <--- Marketing Mind (not my mind)

pretty much

I think to many put stock in a benchmark, not actual game performance.

 

[gururu]

posts get hard to scan through and read with such long posts Rogo and Nutxo; a link would be fine

 

[Blastman]

True you are, but this is a norm in the industry. You really think the Jensen amps put out 1000 watts? No it is an inflated number. It puts out 1000 watts with a THD of 90%.

The THD at Nvidia could be getting a little high these days. LOL

 

[Blastman]

" ...posts get hard to scan through and read with such long posts Rogo and Nutxo; a link would be fine ..

I agree, I couldn't find the post for the quote.

 

[WicKeD]

Originally posted by: Blastman

True you are, but this is a norm in the industry. You really think the Jensen amps put out 1000 watts? No it is an inflated number. It puts out 1000 watts with a THD of 90%.

The THD at Nvidia could be getting a little high these days. LOL

HAHAHAHAH I saw that coming

 

[Blastman]

Man talk about a party hangover from the 5900 launch.

(NV headquarters) ...5900?LAUNCH DAY .. ?..YABBA DABA DO!!! ?. Nvidia back on top!!!.

NEXT DAY???You guys did ? WHAT ? with the benchmarks??

 

[Rogozhin]

sorry bout that long quote I thought it was appropriate to put it in this thread.

rogo

 

[merlocka]

I highly applaud nVidia for this cheat/hack/bug.

Now that people are aware of the types of "optimization" that can be done on a synthetic benchmark like 3Dmark, they have effectively rendered 3Dmark useless as a benchmark.

I may be cynical, but I see this as perfectly fine. If people (consumers or OEMS or Dell,... whoever) are dumb enough to base their purchasing decisions on a synthetic benchmark, then they have to accept this type of crap.

ATI was dumb with the QUACK fiasco, because screenshots could show the difference.

IMO, this cheat/hack/bug from nVidia is very clever and I applaud them for it. People who cry foul from a technical standpoint have completely valid arguments, but the reality is that if you hold a company to the performance of a synthetic (which they don't believe in) this is what you are gonna get. All you can do about it is put your money when your mouth is.

I was gonna switch to a 9700NP this time around but maybe I'll give nVidia another chance just for being so balls-y.

 

[Mem]

I liked what HardOCP had to say,

I am pretty sure you will see many uninformed sites jumping on the news reporting bandwagon today with "NVIDIA Cheating" headlines. Give me a moment to hit this from a different angle.

First off it is heavily rumored that Extremetech is very upset with NVIDIA at the moment as they were excluded from the DOOM3 benchmarks on Monday and that a bit of angst might have precipitated the article at ET, as I was told about their research a while ago. They have made this statement:


We believe nVidia may be unfairly reducing the benchmark workload to increase its score on 3DMark2003. nVidia, as we've stated above, is attributing what we found to a bug in their driver.

Finding a driver bug is one thing, but concluding motive is another.

Conversely, our own Brent Justice found a NVIDIA driver bug last week using our UT2K3 benchmark that slanted the scores heavily towards ATI. Are we to conclude that NVIDIA was unfairly increasing the workload to decrease its UT2K3 score? I have a feeling that Et has some motives of their own that might make a good story.

link

 

[Corporate Thug]

if it in found that nvidia did cheat, i wont buy from them again. knowingly cheating is the same as lying to the consumer. yes ATi cheated ONCE, but they learned from their mistakes.

 

[ProviaFan]

Originally posted by: shady06
if it in found that nvidia did cheat, i wont buy from them again. knowingly cheating is the same as lying to the consumer. yes ATi cheated ONCE, but they learned from their mistakes.

Or has ATI cheated since then? We don't know, and we may never know. Yes, it appears that Nvidia cheated. I know ATI has also, but that won't stop me from buying ATI products, because I like stuff that has decent 3D acceleration combined with good 2D IQ.

 

[waylman]

Originally posted by: Mem
I liked what HardOCP had to say,

I am pretty sure you will see many uninformed sites jumping on the news reporting bandwagon today with "NVIDIA Cheating" headlines. Give me a moment to hit this from a different angle.

First off it is heavily rumored that Extremetech is very upset with NVIDIA at the moment as they were excluded from the DOOM3 benchmarks on Monday and that a bit of angst might have precipitated the article at ET, as I was told about their research a while ago. They have made this statement:


We believe nVidia may be unfairly reducing the benchmark workload to increase its score on 3DMark2003. nVidia, as we've stated above, is attributing what we found to a bug in their driver.

Finding a driver bug is one thing, but concluding motive is another.

Conversely, our own Brent Justice found a NVIDIA driver bug last week using our UT2K3 benchmark that slanted the scores heavily towards ATI. Are we to conclude that NVIDIA was unfairly increasing the workload to decrease its UT2K3 score? I have a feeling that Et has some motives of their own that might make a good story.

link

I've lost all respect for those clowns over at HardON. They are complete sellouts. One day they are kissing ATI's ass, the next NVIDIA's. It's shameful.

 

[Rogozhin]

That was my initial impression of kyle's little blurb, but I am coming to the conclusion that nvidia is going to get flattened by the reviewers after a while.

B3D is really pissed off about the whole thing (not just extremetech).

I will just wait for the smoke to clear.

rogo

 

[nRollo]

LOL

The sky is falling! The sky is falling! nVidia CHEATED on 3dmark!


Some people need to get a grip. There is no right or wrong choice for a VGA. The world won't end if someone buys a different one than you.

nVidia cheated on 3dmark? Big deal. Life goes on. Why do some of you care? Don't buy their cards then.

 

[Mem]

I've lost all respect for those clowns over at HardON. They are complete sellouts. One day they are kissing ATI's ass, the next NVIDIA's. It's shameful.

Lol.... I see your point,however it`s still too early to give a honest verdict,besides it`s not like it effects games or what I consider important software.

In the end only you can decide how important 3Dmark is,I actually don`t rate it at all on the video front.

As I`ve said before this is getting blown out of proportion,in the end it`s really a minor issue that will be sorted one way or the other.

 

[Rogozhin]

I don't, but it's an interesting topic.

Rogo

 

[nRollo]

if it in found that nvidia did cheat, i wont buy from them again. knowingly cheating is the same as lying to the consumer. yes ATi cheated ONCE, but they learned from their mistakes.

Shady:
Of course, be living with VGAs that are worse at some things and paying more on occasion for less performance, but you'll show them!

You don't know ATI cheated "ONCE" you know they got CAUGHT once. If you don't think companies "lie to the consumer", you must be from some other planet where they all don't?

Shady:
You buy the card. You play your games at the settings you like. If they "lied" and it doesn't perform like you'd expect, return it.

 

[Shamrock]

NVidia didnt cheat, merely exposed a flaw in a synthetic benchmark...they openly admitted it about a month ago didnt they? they could manipulate results in 3dmark, and has been outspoken about it since then...they are just PROVING that synthetic benchmarks suck, and purposefully exposing this flaw.

I read somewhere (I cant remember where) that there is about 11 flaws in 3DMark! (if only I was young again and could remember things )

 

[Pete]

Schadenfroh:

we cant trust ati for cheating at quake3, we cant trust nvidia for they cheated at 3dmark2003. the only company we can trust is... well.. TRIDENT!, yeh they have never manipulated benchmarks.

Wrong.

nemesismk2:

The people who wonder why Beyond3D is seen as an anti-nvidia website just need to read your above message. Where are the guys from Nvidia?

They aren't allowed to post there by nVidia itself, according to a regular. Sorry, can't dig up the quote itself, but it was made two or three days ago.

WicKeD:

BTW.. ExtremeTech is just holding a grudge because they were not included with the Doom 3 benchmarks.


P.S I will take Kyle's word any day over Dave Salvator

Why so? Kyle has done a few things that definitely cast a suspect light on his motives. He's taken nV's side quite a few times: He exposed Quack because nVidia tipped him off. He railed against 3DM03 because nV didn't like it. And now he's calling ET's credibility into question b/c they exposed nV's hack. Notice he said is was partly due to jealousy over ET not getting a crack at the D3 benches. I wonder why [ H ] was one of the two sites who got in on that sweet scoop? Notice that Kyle even went so far as to suggest nV did this in order to get caught and expose how poor a benchmark 3DM03 is. Now that's a crock if ever I've heard one. If nV really didn't like 3DM03, they should've prohibited all review sites from benching it with their review cards. They haven't--in fact, they've gone from prohibiting it to allowing it. Now we see why.

As stated Nvidia is not a subscriber to 3DMark, therefore it has no way to view the "floating camera" as B3D did. Since they do not have access to this, they did not see any corrupt images when testing their drivers against it.

Let me ask you this.... Do you have a Nvidia card? If so, are you using the 44.03 drivers? I am using the 44.03 drivers and see no graphic corruptions while I get an increase in performance. So whatever the driver team at Nvidia did, they deserve a thank you. The Nvidia drivers always get better and better.

They didn't need to be in on the Beta to know how their hacks would perform when the camera was taken off its track--everyone seems to be giving nVidia driver engineers so much credit, at least give them credit enough to know how their cheats work. That said, nV was in on the 3DM03 Beta program for at least a few months--no doubt they pulled out when they realized it was going to show their cards in a bad light.

And everyone's drivers get better and better, not just nV's. Note how the 8500 started out behind a GF3, and now usually outperforms the Ti500.

I'm amused in what force the fans are out. Of course the posters in B3D's forums are fans of 3D tech, too, but, unlike most of the people "defending" nVidia in this thread, they actually know what they're talking about.

I can't believe you apologists are spending so much energy defending nVidia's cheating. Why don't you yell at nVidia to spend their time and energy improving their drivers for actual games, rather than wasting their efforts to cheat on a benchmark most of you seem to care so little about?

The reason many more knowledgable people are so upset about this is because nVidia themselves said they didn't like 3DM03 because it could be so easily optimized for. Who was nVidia afraid of--themselves?

merlocka: ALL BENCHMARK DEMOS ARE SYNTHETIC IN THE SAME WAY. You use a pre-recorded demo that essentially takes you on rails through the game. Futuremark can guard against this sort of cheating by changing the "rails" a bit each run, but then that wouldn't allow them to create a huge database of scores for the same run-through.

Mem: Let me present Kyle's post with some highlights of my own that may expose his ostensible third-party impartiality.

I am pretty sure you will see many uninformed sites jumping on the news reporting bandwagon today with "NVIDIA Cheating" headlines. Give me a moment to hit this from a different angle.

First off it is heavily rumored that Extremetech is very upset with NVIDIA at the moment as they were excluded from the DOOM3 benchmarks on Monday and that a bit of angst might have precipitated the article at ET, as I was told about their research a while ago. They have made this statement:


We believe nVidia may be unfairly reducing the benchmark workload to increase its score on 3DMark2003. nVidia, as we've stated above, is attributing what we found to a bug in their driver.

Finding a driver bug is one thing, but concluding motive is another.

Here's Kyle telling us not to conclude motive from repeatable proof of nVidia's driver cheating. Yet he goes on to conclude an evil motive for ET's article based on rumor.

I read [ H ]'s reviews because of Brent, not because of Kyle. Brent also hangs out at B3D's fora and makes an efforts to learn more about the science behind benchmarking, which is why people there tend to trust his reviews--in fact, I started reading his reviews b/c of links and recommendations from B3D. I'm becoming more suspicious of [ H ] again.

All this cheating is just more work for the people who want honest answers, which I'd imagine to be all of us. :\

 

[Schneider]

Who cares if nvidia cheated in 3D Mark 03. Gettin a score of 5700 in a synthetic benchmark doesn't show how fast a
gfx card really is. The fps in games makes a difference.

And there is a massive difference if they were cheatin in game benchmarks like quake 3 etc. The idea is... just as long as a game is :
1) Playable
2) Playable under conditions like 6xAA 8xAF etc
3) Image Quality is good

Those are the three critea u should use when buyin a fast gfx card.
NOT GETTIN HIGH SCORES IN 3D MARK 03

 

[Pete]

Schneider, again, all benchmarks are run the same way--they're all "on rails." Thus, nVidia could optimize the same way for any other game--Quake 3 or UT2K3 or SS2, for instance--and you'd see high scores in reviews, but get low scores when you tried actually playing the game (moving outside the timedemo "rails"). So this issue affects us all, particularly reviewers, who may now have to waste time double-checking benchmarks, devising new timedemos for every review, and re-running new benchmarks on older cards if they want comparative numbers in new reviews (whereas previously they could just pull up numbers from a previous review done on the same system).

I find Anand's Q3 numbers (230fps for all resolutions, even 16x12 4xAA 8xAF, when other sites show lower performance at thosse settings) even more questionable now. Did nVidia include some culling "optimizations" for Q3, too? More work for everyone to find out. I guess you could thank nVidia for that. :roll:

As for who cares if nV cheated in 3DM03? Obviously nV cares, if they took the time to cheat. And you should care, because if they took the time to cheat on 3DM03, what's to say they won't do the same with other timedemos in other benchmarks?

 

[nemesismk2]

Originally posted by: Rogozhin
That was my initial impression of kyle's little blurb, but I am coming to the conclusion that nvidia is going to get flattened by the reviewers after a while.

B3D is really pissed off about the whole thing (not just extremetech).

I will just wait for the smoke to clear.

rogo

Does Beyond3D even have a Geforce FX5900 and if so where is their review?