Is NVIDIA cheating to sell graphic cards?

[Onde Pik]

Wow already way too much for me to read here.. I will just post this.


It is a cheat because the whole "game test" concept disapears from the bencmark when Nvidia exploit the fact that 3DMark uses the same standard flyparth every time.

In normal gaming there is no way for the driver to predict how you are gonna move, so this cheat from nvidia completely eliminates the point of having those gametests atall.

If people can't see this, I am really amazed. But you know what they say, never underestimate the probability of stupidity.

[Behemoth]

Originally posted by Hanners
Link please? I'm kind of confused, because you said you made it up just now.

go to www.cgtalk.com, Technical and Hardware Forum, search "specviewperf cheat"

[SurfMonkey]

If optimising code for specific goals is considered cheating does that mean that all driver optimisations are cheating. After all they all, no matter for what reason, artificially inflate the score somehow.

Should nvidia and ati stop optimising their drivers for games and benchmarks because that is cheating? If the end result is a playable game and the user is happy with what they have in front of them then what's the problem? The user get's what they pay for and the IHV is providing the service for whic they were payed.

I think this is being blown out of proportion and fifty percent of the furore around it can be ignored. If anything it has highlighted the danger in relying on synthetic benchies to provide any truthful result.

[Hanners]

Originally posted by SurfMonkey
If optimising code for specific goals is considered cheating does that mean that all driver optimisations are cheating. After all they all, no matter for what reason, artificially inflate the score somehow.

These are a little more than optimisations though, they are actually blowing away large chunks of the scene being rendered to inflate scores.

The simple fact is that any IHV with scruples wouldn't even consider resorting to this kind of heavy-handed 'optimisation'. Even compared to 'Quack' (which was bad enough) this is taking a huge shortcut, and is basically cheating nVidia's paying customers out of knowing the whole truth about their products.

I'd love to think that nobody buys a video card solely based on 3DMark score, but sadly people do, and these people are being duped by nVidia at the end of the day.

In my eyes, there's no excuse for it - NV35 would have been a great product without resorting to this.

[Hanners]

Originally posted by Behemoth
go to www.cgtalk.com, Technical and Hardware Forum, search "specviewperf cheat"

Just read the thread in question, I think you omitted to post the other side of the story being given out over there:

Originally posted by CgFX
You have totally lost me so I need to bow out. I honestly don't think you know what SPECapc is versus SPECviewperf. They are different.

The current applications used in the five SPECapc benchmarks are Solid Edge v.12, Pro/Engineer 2001, 3ds max 4.2, SolidWorks 2001Plus, a Unigraphics V17.

All of these benchmarks use application specific macros that load data sets and do common user tasks. Our testing with UG has shown SPECapc for Unigraphics to be an accurate generalization on the performance differences between graphics and system solutions.

This is why ATI's SPECviewperf cheating had no baring on their lower than expected SPECapc scores. You pretty much can't cheat SPECapc. You can optimize for it but that means optimizing for the application a particular test is using, and isn't that what we want?

If those aren't applications and typical application usage in your book then I am lost.

[digitalwanderer]

Originally posted by SurfMonkey
If optimising code for specific goals is considered cheating does that mean that all driver optimisations are cheating. After all they all, no matter for what reason, artificially inflate the score somehow.

No, not when they're basing their entire "Our card is the fastest!!!" push around the one benchmark that they intentionally cheated on...then it's called "deception"

I just don't understand this, "They cheated at 3dm2k3, so what?" attitude! If that's the case then why doesn't nVidia step forward and say something about it or put out some non-cheating drivers for an honest comparison?

[DorXtar]

I just finished reading this entire thread.

The best part are the coughcoughnvidiotscoughcough that posted in defense of nvidia before the words "cheating" and "nvidia" were together in headlines all over the intarweb (the last posts were around 1am this morning).

It's a lot quieter now, except for the people who are most disgusted (privately of course) with themselves for forking over so much $$$$ for their nvidia product.

The headline can even be found on the front page of this site (nVnews for those who are lost in this thread).

[SurfMonkey]

Originally posted by Hanners
These are a little more than optimisations though, they are actually blowing away large chunks of the scene being rendered to inflate scores.

The simple fact is that any IHV with scruples wouldn't even consider resorting to this kind of heavy-handed 'optimisation'. Even compared to 'Quack' (which was bad enough) this is taking a huge shortcut, and is basically cheating nVidia's paying customers out of knowing the whole truth about their products.

I'd love to think that nobody buys a video card solely based on 3DMark score, but sadly people do, and these people are being duped by nVidia at the end of the day.

In my eyes, there's no excuse for it - NV35 would have been a great product without resorting to this.

The alternative view is that they are rendering the scene more intelligently than it is currently. By rendering the sky last and only once and also culling (known) useles information in realtion to the view frustrum they bringing an elegant solution to the rendering.

Of course this won't work outside a fixed viewpoint, but in this case it doesn't matter. Nobody except the beta program users will ever see it. If it screwed up games as well then it would be a real nasty issue.

Before I chomped, I do believe that nvidia shouldn't have to resort to such tactics in the first place. But then again neither should ATI, Trident etc... It's only the fact that should a big deal has been made out of 3DMark scores overs the past few years that has forced IHVs do try such methods. The only people who have really gained fom this are FutureMark. But since this calls into question the whole validity of their core business that could soon end.

Like I've said before, this is a bad thig for nvidia but it is probably a good thing in the long run. It is highlighting the fact that our reliance on synthetic benchmarks, as a gauge to performance, severly misplaced and ill conceived. Synthetic benchmarks are just a way for driver teams to show how great they are at optimising for a specific piece of code and how cunning they are at hiding the fact. And that applies to all GPUs.

[Hanners]

Originally posted by SurfMonkey
The alternative view is that they are rendering the scene more intelligently than it is currently. By rendering the sky last and only once and also culling (known) useles information in realtion to the view frustrum they bringing an elegant solution to the rendering.

Elegant?

If I owned a nVidia product, I think I'd be more upset about all the man hours that must have been wasted changing the clip planes in every frame of the benchmark than anything else. Think how much work could have gone into optimising real games and applications, instead of hacking every frame of a benchmark?

[John Reynolds]

Originally posted by Hanners
Elegant?

If I owned a nVidia product, I think I'd be more upset about all the man hours that must have been wasted changing the clip planes in every frame of the benchmark than anything else. Think how much work could have gone into optimising real games and applications, instead of hacking every frame of a benchmark?

EXACTLY!

[gmontem]

I voted yes, but I have a feeling this was also done deliberately to please their stockholders. I'm sure a large number of their investors don't play games at all (they're all in it for the money after all) and a newsflash that NVIDIA beats ATI in 3DMark03 could mean more NVDA stocks bought or held. I wonder what will be said in the next conference call.

[Hanners]

Originally posted by CoWBoY
So if joe blow BIAS 17 year old sales person/H.S. drop out says this card is better, a naive customer is going to say, "Well, they guy at the shop said it was great!" Although, I like this comment better, "My computer guy says this one is the best!" Like that dumbass is telling you the truth or has shown you a comparison of +/-'s about both companies and allows you to decide.

How does that dumbass decide which card is best in the first place though? I would wager 3DMark scores would play a big part in that decision.