ATI Technologies Admits Cheating the Drivers To Achieve Higher 3DMark03 Score

[Rogozhin]

Here is the info.

Looks like what ati though was "optimizing" can be construed as "cheating" and has admitted to the "cheats".

morecheating

What are your thoughts?

rogo

 

[boyRacer]

Matrox pwnz j00!

<--- actually i never owned a matrox card ever... i just find this cheating thing a bit played out lately.

 

[vss1980]

This is old news.... (appeared at Rage3D a while ago, not long after the initial futuremark report)...

And no, I dont class it as cheating myself. An example of optimisation that is a cheat is one which was used by S3 on the Savage2000 so that it would compare favourably in tests against the Geforce where it would not allow a Z-buffer bit-depth higher than 16-bits and hence when other cards used 32-bit colour and increased Z-buffer depth, the Savage only used higher colour depth and hence didn't use extra bandwidth.

ATI's optimisations are exactly the kind of optimisation you see used in the realy world and drivers have been detecting when a certain app is run and optimising the pipeline for years. In many cases its not just for a performance boost, but to fix bugs.

I would class nvidia's failure to use the required floating point precision when rendering as a cheat akin to that of the S3 one mentioned above, and further more demonstrates their inability to properly get WHQL certified drivers for quite a while with the FX cards.

 

[AgaBoogaBoo]

I guess you could consider that "cheating" but if it also makes in overall improvement in games, then it shouldn't be. I haven't tested this but if in the ne xt driver release, people notice a loss of a frame or two per second, then we may know why.

In Nvidia's case, I think its cheating but in ATI's case I don't think its cheating.

 

[Rogozhin]

this isn't old news, it appeared yesterday at xbit and on b3d.

you might be thinking of "ati's response to the futurmark scandal" over at rage3d.com

rogo

 

[MercenaryForHire]

I wondered why the ATI fanboys suddenly shut up about the Detonator tweaks.

Now I know.

/goes off to buy a Matrox

- M4H

 

[DRaG0nBLaD3]

Regardless of how long ago it was first posted, ATi agrees with me . Any modification of code in a synthetic benchmark is cheating. And they're going to take the 'optimizations' out of the next set of drivers, even though there's already a patch that makes them ineffective What more could you ask? ...Well, aside from not cheating in the first place. *shrug* And then there're the matters of scale and method of these cheats....ATi at least did something that could be legitimately done in games, and that didn't make a very significant change in performance.

 

[vss1980]

Oh.... I fail to see what new has been said in this report compared with this at Rage3D..... all xbit and beyond3D have done is basically padded the report out a bit with the respective opinions of the authors.... apart from that this xbit report brings nothing new except being 2 days late to the party.

I doubt these optimisations will be completely discarded and will most probably be worked into the drivers to work alongside other optimisations so that the drivers are best suited to anything you could possibly throw at them. The only change is that rather than an optimisation being initialised by a certain program running, it will be left to the drivers to pick the best method for what its being asked to do.

The only bad thing about ATI's optimisation was that it relied on the detection of a program running so some class it as a cheat, however it does what its asked and cuts no corners.... I fail to see the complaint.....
I fail to see anyone unhappy about the use of compressed textures or Z-compression and colour compression techniques....

 

[SilverTrine]

Tim Sweeney calls what ATi did a legitimate optimization, I believe he knows a little bit more about this sort of thing than you do.

Go and read what he says at www.beyond3d.com.

If you still think ATi cheated after that you can register at www.nvidia.com as a mindles fanboi.

 

[Rogozhin]

vss

I knew this wasn't old news (3days late) but I didn't know it was on the 24th, sorry for calling it out.

I too believe that this isn't a "cheat" but an optimization when looked at within the progamming community, mainly sweeney and teluchin.

I am interesed in everyone's posts though so keep them coming.

rogo

 

[vss1980]

Rogo, not rattling your chain, note its xbit that I criticise.

I think cheating calls are gonna be more difficult to judge in the future, especially considering the amount of freedom allowed for what the card and drivers are doing in terms of DX9 shader instructions (such as exiting functions early if it deems progressing through the whole shader loop is unnecessary).

In this case nvidia's was unbelieveably poor, while ATI's was acceptable. However, I think ATI have bigger problems to solve (like Half-Life escape key usage).

 

[Rogozhin]

vss

thanks

from what I've read over at rage3d isn't the esc bug fixed with the cat 3.4s?

rogo

 

[vss1980]

Yeah, but the 3.4's suffer from massive frame-rate drops in half-life...... so you have to choose an evil..... or do as I do and stick with the 3.1's....

 

[Schadenfroh]

Originally posted by: boyRacer
Matrox pwnz j00!

<--- actually i never owned a matrox card ever... i just find this cheating thing a bit played out lately.

Schadenfroh<---- cant wait for parhelia 2

 

[bunnyfubbles]

frankly I don't care as long as I can run my Tribes 2 and other games with respectable performance and visual IQ.

Oh and I'm holding out for the Parhelia 3

 

[gururu]

i think it is pathetic that ati sees a need to posture at this point in time. nvidia is on the grill, and any statements from ati only obscure the issue. what are they trying to do? show up nvidia?
it fuels the ati/nvidia debate and divides the end user base. really pathetic.

i'm almost....almost inclined to look over the nvidia cheats.

and oh by the way, i feel this way because i am not convinced that atis optimizations qualified as a cheat.
therefore... yeah i think ati is posturing.

 

[Pete]

How about John Carmack's opinion?

Allow me to highlight the part specifically relating to nVidia's (first paragraph) and ATi's (second paragraph) respective roles in the matter:

Rewriting shaders behind an application's back in a way that changes the output under non-controlled circumstances is absolutely, positively wrong and indefensible.

Rewriting a shader so that it does exactly the same thing, but in a more efficient way, is generally acceptable compiler optimization, but there is a range of defensibility from completely generic instruction scheduling that helps almost everyone, to exact shader comparisons that only help one specific application. Full shader comparisons are morally grungy, but not deeply evil.

 

[Rogozhin]

thanks for the link pete, interesting reading.

rogo

 

[gpw11]

This reminds me of the whole "ATI cheating on Q3 time demos" thing from a while back.

 

[nRollo]

Silvertrine:

If you still think ATi cheated after that you can register at www.nvidia.com as a mindles fanboi.

I'm starting to think you should change your name to "One Hit Wonder"- apparently all you have to say is "fanboy".

As far as the rest goes, I have been saying this for days, and I don't care if ATI cheats at 3dmark either. I'll still buy their cards. (even though I'm a "fanboy"

)

As far as "ethics" goes, someone will have to point me at something more on the level of child slave labor to get me to stop buying ATI or nVidia cards. I like to buy new VGAs. I'm not going to cut myself off from one of the only two companies who build good VGAs because they cheat at benchmarking. Who knows? One of these companies might build something that is significantly faster than the other, then I would have to deny myself the card or look like an idiot because people might remember me ranting,"I will NEVER buy ATI/nVidia! They CHEATED on 3dmark! THAT is unexcusable!".

So, reviewers can point fingers, BBSers can whine about "not fair", but I'm smart enough to know I only have two options, and if they've both been caught cheating before, I'm not going back to my V5 so I can pretend I care about trivial BS like 3dmark.

 

[AdvancedRobotics]

Originally posted by: Schadenfroh

Originally posted by: boyRacer
Matrox pwnz j00!

<--- actually i never owned a matrox card ever... i just find this cheating thing a bit played out lately.

Schadenfroh<---- cant wait for parhelia 2

Me neither! I have the parhelia and all I have to say is Matrox basically owns! Best graphics card company, IMO

--aR Labs

 

[Rogozhin]

I'm hunting FXs tonight.

rogo