View Full Version : Is NVIDIA cheating to sell graphic cards?
Originally posted by Nv40
Hanners.. in what way is diferent SPECviewperf from 3dmark2003 then ?
in 3dmark 2003 scores its not going to translate to any games
have you seen DOOm3 scores? have you seen SPlinter cell scores?
UNrealT2003 scores? Quake3?
3dmark2003 *is* and Unfair benchmark and not represntative
of any future in games ,except 3dmarks . because no game will use that Engine. IMO.. anyone who doesnt want to see this is in complete denial of the Truth.
if i remember well ..
Game1 -> a DIrectx7 game
Game2/3/ ->Pixel shadels 1.4
Game4 ->pixel shaders 1.4 at near everything /only PS2.0 for the sky.
the benchamrk is biased too cards with powerfull Vertex shaders..
coincidentially here is where ATI clearly leads NVidia ,benefits more cards ,very fast in single texturing ,again ATI scores .
and if that where not enough it benefit greatly cards with 8x1 designs in their pipes.. so much coincidence , and as a Bonus it force Nvidia Nv3x cards with nothing more than 32bits more of greater presision than ATI ones.. 128bits vs 96bits... too much presision for so mediocre sky.. in Nature test ,that looks to me a 2d picture.. is that FAIR?
wait i forgot another one.. ATI have access to the source code ,
and they have also the advantage here ,because they are members.
did i need to say more??
what would you say if Nvidia were memebers of a benchamk
and not ATI?
did you really believe Nvidia is the Only one company that cut corners
to gain free performance ? the only one with misleading marketing?
so maybe Nvidia was unfair too , so what? the benhmark is unfair
since the begining. the only diference between an Mpeg movie
and 3dmark2003 is that the later give you misleading scores..
of a performance you will never see in your games ..that all.
what i find more deplorable is not this 3dmark unfairness ,
but the People that clearly understand this ,and still point their
fingers to Nvidia for being Unfair ... in an Unfair benchmark. [/B]
to continually question futuremark in this matter over and above the fact that nvidia obviously did something that did not conform with the guidelines of the benchmark is not right IMO...
it is a matter of opinion whether or not 3dmark03 is a good benchmark or not... IMO perhaps it could have been better... no doubt about it... BUT it is recognized as stressing the video card and it DOES give a representative score for the products as they perform...
notice the difference in performance between ati and nvidia products... there does not appear to be that big a difference with some of the older drivers (precision is called into question but lets let that slip for the sake of argument)
AFAIK nvidia chose to leave on their own from the development of 3dmark03 @ the beta stage...
they then attacked the product released for various reasons... all of which were answered in the white paper... but here is the problem...
nvidia claims that 3dmark03 is a faulty benchmark... fair enough... they then go ahead and have 'optimised' for 3dmark03 since the release of the nv30 and have now very blatantly been caught 'optimizing' or rather cheating in the tests...
the tests are designed to run a particular way... if you 'optimise' in order to find shortcuts in the benchmark you are circumventing the program parameters and hance... are cheating...
the validity of 3dmark03 cannot be raised time and again... the logic that because 3dmark03 is not representative in the eyes of some people... that it automatically gives nvidia (which has professed its dislike very publicly) the right to CHEAT in that benchmark... is inherently flawed...
also... if nvidia is pulling the same stunt in benches such as SS:SE or ut2k3... what then is the view ?
concerning 3dmark03 nv40... I was distinctly under the impression the precision was between int12 and fp16 for nvidia v/s fp24 for ati... the dx9 spec calls for fp24 precision which ati r3xx cards do all the time... for gametest 4...
thats what I thought anyway :)
Blomman00
05-15-03, 04:10 PM
Driver bug, case closed ;)
Originally posted by Blomman00
Driver bug, case closed ;)
Exactly! NVidia's perfect drivers accidentally detects a certain scene in 3DMark03 and then accidentally does not clear the frame buffer. This sort of accidental scene detection is a very common occurance in drivers.
Originally posted by Nv40
indeed .
and that is that people dont play games anymore ,
but rather spend their time hunting for lost pixels in 3dmarks ,
just to find another "flaw" in the GeforceFx.
if extremetech claims are accurate, it only proof that Nvidia
is cutting curners in that benchmark , it is unfair for ATI ,so what?
3dmark2003 is unfair for Nvidia cards anyway ,since day 1 .
thats why ,they dropped from the beta team , while ATi not.
ATI and Nvidia have diferent arquitectures ,so its not possible
apples vs apples comparisons when it comes to PS/VS.
because everyone doeas things diferently..
so you dont need to be a rocket scientist to know syntetic
benchmarks can be unfair for other companies.
ask AMD ,they will tell you ,many stories about this.
Is Nvidia cheating to sell "graphic cards?"
as far as i know all video cards companies have done this..
ATI,Trident,and according with ET. Nvidia.
any company will take every opportunity they have
to cut corners here and there to gain Free performance.
(look at my tread of anisofiltering )
but they sell graphics cards to earn money ,not to "cheat". :)
Please tell me why you seem to KNOW that 3DMark is "Unfair" to NV?
I can't wait to hear this one..... they took THEMSELVES out of it. Why exactly was that anyways?
DorXtar
05-15-03, 04:28 PM
Buyer error, case closed ;)
Originally posted by Blomman00
Driver bug, case closed ;)
Iraqi Information Minister
-The fanATIc infidels are committing suicide by the hundereds outside nVidias office in Santa Clara. Our initial assasment is that they will all lie before they die.
Hanners
05-15-03, 04:46 PM
Originally posted by Nv40
good question..
it cost more being a member of FUture mark.. by FAR .
because they not only need to pay for membership ,
but also design their Future hardware Nv3x,Nv4x,Nv5x..
with more transistors and spend hundreds of hours
in drivers tweaking ,just for a useless benchmark.
Isn't this whole thread about the hundreds of hours nVidia have spent tweaking drivers for a 'useless' benchmark?
Hanners
05-15-03, 04:49 PM
Originally posted by Nv40
hanners.. in what way is diferent SPECviewperf from 3dmark2003 then ?
3DMark2003 - A DirectX 9-based benchmark designed to simluate real-world gaming situations, which doesn't use any specific engines that you would find in games.
SPECviewperf - A benchmark for users of professional 3D applications, which uses those very same professional applications in it's tests.
Two very different kinds of benchmarks if you ask me.
reever2
05-15-03, 04:51 PM
Umm you cannot say 3dmark03 is a useless benchmark after everybody went completely ape**** over the prelim 5900 benchmarks which showed it quite ahead in 3dmark03, predicting how a card will do based on a "useless" benchmark is so hypocritical its not even funny
Why would Nvidia tell people that a benchmarking program is not recommended for use with their cards until issues are resolved, and then try to make it seem like their cards rule by cheating with the results of the benchmark program which people shouldn't be using in the first place because it is inaccurate?
Hanners
05-15-03, 05:04 PM
Regardless of people's opinions of 3DMark, I don't see how regarding it as useless or a lazy way to test a card has anything to do with the issue at hand, it doesn't change the simple fact that nVidia cheated on a widely-used benchmark, and has set a dangerous precendent for this kind of thing in any kind of benchmark or timedemo.
Hanners
05-15-03, 05:13 PM
Originally posted by CoWBoY
Doubt they will concern themselves with this in the next driver versions. Maybe we will see some more B.S. with Det 50.xx, although it would be stupid to continue on with it as it is a waste of time/resources.
I hope they concern themselves with fixing this 'bug' before they forget all about 3DMark...
Man Cowboy, ever since you changed that Avatar of yours I see a completely different guy.. and TBO I liked the other guy better;)
Hanners
05-15-03, 05:28 PM
Originally posted by CoWBoY
What would be the point? Who needs justification other than ATi owners?
If nVidia are going to pretend this is a bug, then they should at least make it look as though they are trying to fix it.
The longer they leave it untouched, the further the 'nVidia are cheats' reputation will spread, and the last thing nVidia need right now is to be dragged through the mus over and over again. The sooner it's fixed, the more chance of it being forgotten.
Hanners
05-15-03, 05:38 PM
Originally posted by CoWBoY
When the Det 50.xx come out, I suspect whatever will be will be. Although I wouldn't waste time on it if I really felt that it was worthless. I don't like the idea of Nvidia wasting time on it, if anything, I think it should get worse so you can't benchmark your Nvidia card with it at all. Like a "Quack" with tremendously adverse results not in favor of Nvidia.
I imagine it'll be quicker to remove than it was to put in. ;)
Again, breaking 3DMark totally on nVidia cards would set a dangerous precedent. What next? nVidia don't like the way Half-Life 2 runs on their cards, so they deliberately break the game in the drivers? It won't happen obviously, but the way nVidia are thinking these days I sometimes wonder.
To start breaking applications in drivers would be childish in the extreme, and certainly the final nail in the in the coffin that would lay the 'superior nVidia drivers' ghost to rest.
CaptNKILL
05-15-03, 05:44 PM
Originally posted by CoWBoY
When the Det 50.xx come out, I suspect whatever will be will be.
Jesus, I hope so... if whatever will be DOES'NT be, then we have bigger problems than driver cheating!!!
Sorry, I had to say it :rofl
(ignore me.. im bored)
StealthHawk
05-15-03, 06:48 PM
Originally posted by CaptNKILL
easy there killer.... I said I didnt read the article.... if you read my post you wouldnt be wasting your keyboard life flaming me.
At the time of posting, I didnt know what this was about, I was simply saying "If I cant SEEEEEEEEE the disadvantages of a cheat, I dont have a problem with it". OBVIOUSLY, this cheat involves something that we can SEEEEEEEEEEEEEE so my comment doesnt apply to this particular "bug" does it?
I was refering to the "cheats" people whine about that have no visible disadvantages until you take a screenshot and zoom in 100x. Those kinds of "cheats" dont ruin the picture do they? No, but lots of people complain about them just as much as this cheat.
Dont call me a n00b (or a Nub or whatever this week's stupid word is) if you are just going to skim over my post and rant about the first thing that pops into your head.
But these cheats are NOT something you can see. That's the point. You only know about them if you are a Futuremark beta partner and have access to the beta build with debug features like free look camera.
If the debug version didn't exist no one besides nvidia would ever know or be able to prove such cheats were taking place.
As a side note, with the exception of GT1, 3dmark03 is an extremely poor indicator of current gameplay. It's supposed to represent the possibility of future gameplay. Of course I question whether it really does or not.
ChrisRay
05-15-03, 06:57 PM
Originally posted by Nemesis77
Funny, I don't see anyone whining about it though... I bet that if NV said "we don't think you should use CC because....", it would take about 5 minutes for these forums (and others) to be filled by NV-fans who tell everyone why CC sucks.
"When NV says "jump", only thing we will ask is "how high?""
Again, people were OK with 3DMark as long as it was OK with NV. The second NV complained about it, people started to complain as well.
And how come 3DMark doesn't have any credibility? There are plenty of people using it to benchmark their system, including professional reviewers. Or did it lose it credibility because NV doesn't like it?
You're taking my opinion of 3dmark based on What Nvidia is says, 3dmark2001 and 3dmark2003 lost all credibility with me when I saw ATI And Nvidia spending their time boosting 3dmark scores, Rather than fixing game issues.
3dmark was once a useful tool for testing Overall system performance and comparing systems to other systems. But thats about it. And 3dmark2003 even further damaged it's usefulness in that respect.
Being a primarly GPU dependent Benchmark. It's not really valid for it's over all usefulness of comparing system performance.
Now it's just a bench that benches 3d Graphics cards, Which doesn't reflect any real world game performance, 3dmark never reflected real world game performance. But it's over all value as a tool,
Many we4bsites have stated that they are not sure about the validity of the benchmark, Extreme tech, Toms hardware. Hardocp, And I think anandtech but don't quote me on that.
Anywho, 3dmark's peformance is just what it shows. a Number, It does not reflect real game performance and I doubt it ever will. None of them ever have.
3dmark has too much credibility, Far more than it deserves, Expecially considering it's overall usefulness has been even more degraded since the relese of 3dmark2003,
Now you have a strict GPU bench that does not reflect real game performance, Rather than an overall system bench (3dmark2001) that does not reflect system performance.
But the actual usefulness of comparing systems has been hurt by the release of 3dmark2003 currently.
PsychoSy
05-15-03, 08:21 PM
Since the other thread was closed due to a number people who thought it was a radical act to come on over and "neener friggin' neener" back and forth just to read their own BS back to themselves, I wasn't able to chime in...:mad:
But I'm gonna chime now...and briefly!! :firedevil
Are you people honestly suggesting (some of you downright claiming) that Nvidia purposefully built these drivers to somehow detect when 3DMark2003 is launched, thus implementing these culling planes, resulting in a cheat that would vastly improve the scores in a effort to falsely hornswaggle people into buying more nVidia cards?!?
Am I reading that right? If so, then I think all those people who are claiming "CHEAT!!" are spinning their wheels!! If 3DMark2003 is the ONLY application where these extra culling planes occur, I have a suggestion that could immediately confirm or deny this "we cooked our drivers to detect 3DMark2K3" theory that some of your folks are preaching.
RENAME THE 3DMARK2003 ".EXE" FILE!!!
I'd suggest a name of "thisuselessbenchygivesmearise.exe"!!
Anyway, after renaming 3DMark2K3, run the sucker. If this "bug" is still there and scores remain unchanged, then the "cooked driver" theory is baseless and without merit. But if this issue doesn't crop up and scores plument, then it's clear that nVidia has some serious explaining to do.
Personally, I wouldn't be surprized if nVidia done this just to show how irrelvant 3DMark2003 really is. I'm sorry, but when you have a synthetic benchmark where the same camera-fixed scenes are rendered each time you launch it, then you got something that can easily be manipulated. Any benchmark that uses a "fixed camera" in the sense that the camera's motion is identical every run is pretty much worthless because you can't orient the player 45, 90 or 180 degrees and run the benchmark again.
When you know exactly what's going to be rendered on a screen at every given moment, it gives you all the tools you need to brew up ways of implementing optimization methods that you wouldn't otherwise have, resulting in something that can be easily manipulated. It's an inherant flaw in any kind of "fixed camera" based benchmark and always will be.
What you need is a benchmarking suite that's so sophisticated that it can randomly generate scenes to render in order to test all the different aspects of the hardware's features. Something like the Random Mission Generator in SOF2 but much much more sophisticated and unpredictable. Add some interaction and you've got the Mother Of All Benchmarks because it would TRULLY measure real-world gaming performance unlike what 3DMark or a timedemo can do.
Such a benchmark will never exist.
Both nVidia and ATi would complain that the pure randomness and unpredictability of a such an application would severely hamper their abilities to implement driver optimizations to one up each other. And without their deep coffers funding the project, it would be way too costly for anyone else to develope independantly.
saturnotaku
05-15-03, 08:29 PM
Originally posted by PsychoSy
Personally, I wouldn't be surprized if nVidia done this just to show how irrelvant 3DMark2003 really is. I'm sorry, but when you have a synthetic benchmark where the same camera-fixed scenes are rendered each time you launch it, then you got something that can easily be manipulated. Any benchmark that uses a "fixed camera" in the sense that the camera's motion is identical every run is pretty much worthless because you can't orient the player 45, 90 or 180 degrees and run the benchmark again.
Damn, I was expecting this from the NVIDIA PR machine, not from you bud. :)
Ain't sayin' I agree with 'ya necessarily. But I'm gonna keep my big yapper shut.
Rogozhin
05-15-03, 08:29 PM
So you're saying there are only two options here.
1.You are more intelligent than the folks over at extremetech (and B3D) and can figure out how to prove that nvidia didn't cheat.
2.And two, even if they did the only motivation was to prove the vapidness of 3dmark2003.
I'd have to go with three.
3.You really are a psycho (they don't use logic ;))
rogo
digitalwanderer
05-15-03, 08:32 PM
Originally posted by PsychoSy
Since the other thread was closed due to a number people who thought it was a radical act to come on over and "neener friggin' neener" back and forth just to read their own BS back to themselves, I wasn't able to chime in...:mad:
But I'm gonna chime now...and briefly!! :firedevil
Woo-hoo! It's about time, where ya been? ;)
Are you people honestly suggesting (some of you downright claiming) that Nvidia purposefully built these drivers to somehow detect when 3DMark2003 is launched, thus implementing these culling planes, resulting in a cheat that would vastly improve the scores in a effort to falsely hornswaggle people into buying more nVidia cards?!?
Sort of...yes! :)
Am I reading that right?
Yup.
If so, then I think all those people who are claiming "CHEAT!!" are spinning their wheels!! If 3DMark2003 is the ONLY application where these extra culling planes occur, I have a suggestion that could immediately confirm or deny this "we cooked our drivers to detect 3DMark2K3" theory that some of your folks are preaching.
RENAME THE 3DMARK2003 ".EXE" FILE!!!
I totally agree with that, I wish I had an FX to test it with. :(
Personally, I wouldn't be surprized if nVidia done this just to show how irrelvant 3DMark2003 really is.
I can see that argument, but their heralding of it as their proclamation as the "King" sort kills that idea off....'cept for "post-discovery damage control" ;)
Why are we still kicking this dead horse? Hasn't nVidia given us anything new to chew on with it?
Originally posted by PsychoSy
Since the other thread was closed due to a number people who thought it was a radical act to come on over and "neener friggin' neener" back and forth just to read their own BS back to themselves, I wasn't able to chime in...:mad:
But I'm gonna chime now...and briefly!! :firedevil
Are you people honestly suggesting (some of you downright claiming) that Nvidia purposefully built these drivers to somehow detect when 3DMark2003 is launched, thus implementing these culling planes, resulting in a cheat that would vastly improve the scores in a effort to falsely hornswaggle people into buying more nVidia cards?!?
Am I reading that right? If so, then I think all those people who are claiming "CHEAT!!" are spinning their wheels!! If 3DMark2003 is the ONLY application where these extra culling planes occur, I have a suggestion that could immediately confirm or deny this "we cooked our drivers to detect 3DMark2K3" theory that some of your folks are preaching.
RENAME THE 3DMARK2003 ".EXE" FILE!!!
I'd suggest a name of "thisuselessbenchygivesmearise.exe"!!
Anyway, after renaming 3DMark2K3, run the sucker. If this "bug" is still there and scores remain unchanged, then the "cooked driver" theory is baseless and without merit. But if this issue doesn't crop up and scores plument, then it's clear that nVidia has some serious explaining to do.
Personally, I wouldn't be surprized if nVidia done this just to show how irrelvant 3DMark2003 really is. I'm sorry, but when you have a synthetic benchmark where the same camera-fixed scenes are rendered each time you launch it, then you got something that can easily be manipulated. Any benchmark that uses a "fixed camera" in the sense that the camera's motion is identical every run is pretty much worthless because you can't orient the player 45, 90 or 180 degrees and run the benchmark again.
When you know exactly what's going to be rendered on a screen at every given moment, it gives you all the tools you need to brew up ways of implementing optimization methods that you wouldn't otherwise have, resulting in something that can be easily manipulated. It's an inherant flaw in any kind of "fixed camera" based benchmark and always will be.
What you need is a benchmarking suite that's so sophisticated that it can randomly generate scenes to render in order to test all the different aspects of the hardware's features. Something like the Random Mission Generator in SOF2 but much much more sophisticated and unpredictable. Add some interaction and you've got the Mother Of All Benchmarks because it would TRULLY measure real-world gaming performance unlike what 3DMark or a timedemo can do.
Such a benchmark will never exist.
Both nVidia and ATi would complain that the pure randomness and unpredictability of a such an application would severely hamper their abilities to implement driver optimizations to one up each other. And without their deep coffers funding the project, it would be way too costly for anyone else to develope independantly.
That's pretty stupid as nobody is claiming it is detecting the filename. There are many ways to detect if a certain application is running that does not involve checking the filename.
Your claim that a benchmark with a fixed camers is worthless is stupid also. This would mean Quake3 and vitually all other benchmarks are irrelevant also.
Do you get your talking points straight from nVidia?
I beg to differ on the irrelevancy issue.....
If anyone thinks that the scores mean nothing to consumers they are just fooling themselves, money is made off of that, and there is NO denying it.
There are some folks who see whats going on, and those folks will not bite..... but there are MORE folks who just see the #'s and BUY!!!!... Without coming to forums like this and B3D to find out the true stories behind the benches...... all they see are scorching scores, because they don't know any better.
These are my views on it, I'm sure there are others.
All ya really gotta do to prove what I am saying is this:
Think back to when you were a noob...... you knew NOTTA/SQUAT!!!.... and half ( or more- probably alot more) of the folks that buy high end hardware know exactly that.
I am CERTAIN there are plenty of folks who know AT LEAST one person who buys " nothing but the best"...funny thing is they don't know jack, they read magazines or flip through benchmarks without digging for info ( they figure why would anyone lie)...
Then there are the types who listen to what the biased ( or just very uninformed for typical reasons) salesman telling them such-n-such is WAY better........
Half of them clowns know jack also.
vBulletin® v3.7.1, Copyright ©2000-2012, Jelsoft Enterprises Ltd.