NVIDIA Detonator 51.75 Driver Performance

[Gorgon]

For the past couple of months or so, during my different meetings with NVIDIA at trade events or on their campus in Santa Clara, I have been hearing the words of the Release 50 Detonator drivers. They were to offer dramatic performance increases as well as new features that would definitely grab the eye.

Well, after todays events between ATI, NVIDIA and Valve, NVIDIA sent over their release 50 drivers, version 51.75, to be tested.

Now, while I don’t have a copy of the new beta Half Life 2 to benchmark using these new 51.75 drivers, it is these drivers that will give the NVIDIA graphics cards the much needed push in performance in DX9 games, including HL2. While we currently don't have good ways of testing DirectX 9, I decided it would be good to see how the new drivers affect current games running the DX8 API..

Below you will see benchmarks comparing the 44.67 drivers to the 51.75 drivers on an NVIDIA GeForce FX 5900 Ultra graphics card. Driver settings are shown on the graphs as well.

click her please for the rest:
http://www.amdmb.com/article-display.php?ArticleID=257


:cheers:

 

[wormstrangler]

Can we see some Dx9 performace from Tomb Raider or something? Because I doubt most people are interested in Dx8 performance anymore.

 

[scribblehead]

Well its a step in the right direction, i would still wait to see the drivers tested in the HL2 benchmark before deciding what card to buy---but i'm almost 100% confident that i'm getting a radeon.

 

[TripleDES]

Weird how in the 3DMark03 bench the PS2 performance made a jump. I wonder what the driver does internally...

 

[FluxCap]

Doh, not good.

 

[Jerjerod]

Cant you test it on tomb raider or the halo beta?

 

[SLH]

lol, look at the last graph, one of 'em went down!

 

[dawdler]

Need to see it in a game... And actually all the games with the slightest hint of DX9 we got to make an accurate guess if it really is improvement or just haxed drivers. Going my 3Dmark isnt a good idea. Going just by TRAOD or Halo isnt a good idea either, they could have "optimised" them by now.

 

[snark^]

Beyond3d did some benchmarking with TRAOD recently. It wasn't pretty for nvidia there either. Strange that a TR game would be a source for benching, but as explained in the article it is the first game to fully use DX9 features.

http://www.beyond3d.com/misc/traod_dx9perf/

Machine: P4 3GHz, 1GB DDR RAM, DX9.0a, 1024x768x32 @ 60Hz : FPS
GeForce FX 5200 Ultra : 6.83
GeForce FX 5600 Ultra : 8.41
GeForce FX 5800 Ultra : 18.76
Radeon 9500 : 21.48
GeForce FX 5900 Ultra : 23.19
Radeon 9600 PRO : 29.67
Radeon 9700 PRO : 49.62
Radeon 9800 PRO : 51.52

 

[dawdler]

Originally posted by snark^
Beyond3d did some benchmarking with TRAOD recently. It wasn't pretty for nvidia either. Strange that a TR game would be a source for benching, but as explained in the article it is the first game to fully use DX9 features.

http://www.beyond3d.com/misc/traod_dx9perf/

Machine: P4 3GHz, 1GB DDR RAM, DX9.0a, 1024x768x32 @ 60Hz : FPS
GeForce FX 5200 Ultra : 6.83
GeForce FX 5600 Ultra : 8.41
GeForce FX 5800 Ultra : 18.76
Radeon 9500 : 21.48
GeForce FX 5900 Ultra : 23.19
Radeon 9600 PRO : 29.67
Radeon 9700 PRO : 49.62
Radeon 9800 PRO : 51.52

Yes, but that wasnt using the new drivers, which is kind of the point with the entire thread

 

[snark^]

It's not going to make that much of a difference though. It appears the best they can hope for is to make the framerates playable.

ATi is yet to release the cat 3.8's yet too and they're being hyped up as another huge step for gaming-kind. HL2 had better be worth all the hype and trouble it's causing the industry...


PS: For those who can't be bothered reading that benchmarking article the fps chart I quoted above were obtained with Cat3.6 and Det 45.23 drivers BTW.

 

[DimitriPopov]

Update (9/12/03 12:52am): Interestingly, in Anand's article addressing the HL2 performance issue, Valve specifically gave instructions to NOT use the Release 50 drivers which seems rather odd to me. If they could improve HL2's performance, then they would want that to be known, right?

Hmm thats STRANGE isnt it. How very suprising. /sarcastic tone

 

[Gorgon]

dam, why is that?

can you email gabe about this?

 

[Pr()ZaC]

LOL!
Do you really think that nVidia will automagically fix their drivers to improve speeds THAT much without IQ reduction?
Sorry!

The sacrifices you encounter by running either the mixed mode path or the DX8 path are obviously visual. The 5900 Ultra, running in mixed mode, will exhibit some banding effects as a result of a loss in precision (FP16 vs. FP32) but still looks good - just not as good as the full DX9 code path. There is a noticeable difference between this mixed mode and the dx82 mode, as well as the straight DX8 path - for example, you'll notice that shader effects on the water aren't as impressive as they are in the native DX9 path.

Dets 50 will descrease shader precision (lower IQ) and will carry a sh!tload of special "optimizations", that's why you'll see improved performances.

 

[mrchimp]

I put 2 turds and a mouldy grape fruite on the det 50 drivers increase HL2's performance by 10fps.

 

[Typhon]

there's no point in all this negativity. no one here has any reason to hope the new drivers WON'T help, so why all the bitching?

detonator drivers have, historically, done some amazing things. There have been several releases that improved performance 50%!!! So I'll wait to see the DX9 results.

 

[dscowboy]

Originally posted by DimitriPopov
Hmm thats STRANGE isnt it. How very suprising. /sarcastic tone

*sigh* If you read Gabe's benchmark presentation, you'd know that he already addressed this. Valve claims that NVIDIA is putting 'features' in their 50 beta drivers that they never intend to release, and that only improve performance specifically in the HL2 bnechmarks, and would do nothing for performance in the actual game. Valve is trying to prevent NVIDIA from cheating on benchmarks AGAIN by forbidding hardware sites from testing HL2 with the rediculous hacks NVIDIA put in their beta drivers.

I'm much more inclined to believe a game developer who wants as many people as possible to be able to run their game as well as possible, than a hardware vendor who has already proven they're quite willing to blatantly cheat on benchmarks. Whether the 50 drivers actually do have changes that will vastly improve DX9 performance or not, It's pretty damn disgusting that NVIDIA would take the benchmark Valve provided them to help them improve performance, and instead use it to design camera-path specific hacks to inflate their numbers.

 

[dawdler]

Originally posted by dscowboy
Valve is trying to prevent NVIDIA from cheating on benchmarks AGAIN by forbidding hardware sites from testing HL2 with the rediculous hacks NVIDIA put in their beta drivers.

I think that is only about 10% of the reason.
The main reason is that they are testing it with CURRENTLY AVAILABLE DRIVERS FROM BOTH COMPANIES. Nothing more. Nothing less. This is as fair as it can be. No unnofficial beta Nvidia driver against the officially released Radeon. This gives people the results of it if they would run it TODAY. People cant run with the 51.75 or whatever driver Nvidia want them to use.

Simple as that. No conspiracy theories on cheating.

Also, why should Nvidia be selective? Without a doubt, ATI is including performance optimisation for HL2 in the 3.8 series. I would be surprised if they didnt (me hoping it will take the form of increased DX9 performance across the board). You dont see them complaining that a specially engineered beta 3.75 isnt being used. Of course, they dont need it, but that's not the point.

 

[dscowboy]

Originally posted by dawdler

Also, why should Nvidia be selective? Without a doubt, ATI is including performance optimisation for HL2 in the 3.8 series.

Read Gabe's slide. He accuses NVIDIA not of including HL2-specific features, but of including BENCHMARK-SPECIFIC features. Features that would only improve performance in the benchmark, NOT in any actual game. By doing that, NVIDIA is undermining the whole concept of benchmarks. What's the point in HAVING benchmarks if they don't reflect in-game performance?

And not only that, but they're allegedly doing things like degrading the image quality but turning all the features back on briefly if the user tries to take a screenshot! If this is true there would be no way that hardware sites could post screenshots to illustrate the image quality discrepancy between ATI and NVIDIA.

According to Gabe's presentation, this is not just about two hardware manufacturers trying to add genuine optimizations to their drivers, it's about one manufacturer trying to cheat and mislead people.

 

[mrchimp]

Originally posted by dscowboy

And not only that, but they're allegedly doing things like degrading the image quality but turning all the features back on briefly if the user tries to take a screenshot! If this is true there would be no way that hardware sites could post screenshots to illustrate the image quality discrepancy between ATI and NVIDIA.

I'm not sure if even Nvidia would try and do something like that, although if Gabe said they have then it's proberly true.

There's an easy way to get round it if they have done that though, quite simply take a picture of the monitor whith a hi res camera. As long as they get the lighting conditions right it should indicate whether Nvidia are indeed cutting corners, and possibly setting themselves up for a law suit.

 

[dawdler]

Originally posted by dscowboy
Read Gabe's slide. He accuses NVIDIA not of including HL2-specific features, but of including BENCHMARK-SPECIFIC features. Features that would only improve performance in the benchmark, NOT in any actual game. By doing that, NVIDIA is undermining the whole concept of benchmarks. What's the point in HAVING benchmarks if they don't reflect in-game performance?

And not only that, but they're allegedly doing things like degrading the image quality but turning all the features back on briefly if the user tries to take a screenshot! If this is true there would be no way that hardware sites could post screenshots to illustrate the image quality discrepancy between ATI and NVIDIA.

According to Gabe's presentation, this is not just about two hardware manufacturers trying to add genuine optimizations to their drivers, it's about one manufacturer trying to cheat and mislead people.

Yes, but as far as I have read, he havent confirmed that the newest 50 drivers have all these "optimisations". But true, he is accusing them of alot
And its about damn time. Futuremark is pointless now. No one trusts it.
And if this screenshot "bug" is true (no, Nvidia would neeeeeeeever include such a thing, of course, horrible thought) it is really scary... really really REALLY scary... Cause it means that reveiwers cant trust anything. EVERYTHING we have seen could be false. If they can prove this is true, I will never trust Nvidia again. Maybe, possibly, if they admit it when its proved, and give a damn good "I'm sorry" text from the director.
Clipping planes are at least just speed increases, although wrong. Replacing shaders is also wrong, but if it produces nearly the same results, its not outright horrible, just very wrong.

Manipulating a screenshot is straight out deception. I equal that to a rape.

 

[Heleos]

Why would they do that? They would be asking to be caught, because someone who wanted to prove it could do it really easily... it's not like nVidia could get away with something like this, so why would they do it?

 

[dscowboy]

Originally posted by Heleos
Why would they do that? They would be asking to be caught, because someone who wanted to prove it could do it really easily... it's not like nVidia could get away with something like this, so why would they do it?

They DIDN'T get away with it, Gabe told every major hardware review site about it!

 

[TrueWeltall]

Originally posted by Heleos
it's not like nVidia could get away with something like this, so why would they do it?

To make money.

 

[Blair]

Originally posted by G0rgon
For the past couple of months or so, during my different meetings with NVIDIA at trade events or on their campus in Santa Clara, I have been hearing the words of the Release 50 Detonator drivers. They were to offer dramatic performance increases as well as new features that would definitely grab the eye.

Well, after todays events between ATI, NVIDIA and Valve, NVIDIA sent over their release 50 drivers, version 51.75, to be tested.

Now, while I don’t have a copy of the new beta Half Life 2 to benchmark using these new 51.75 drivers, it is these drivers that will give the NVIDIA graphics cards the much needed push in performance in DX9 games, including HL2. While we currently don't have good ways of testing DirectX 9, I decided it would be good to see how the new drivers affect current games running the DX8 API..

Below you will see benchmarks comparing the 44.67 drivers to the 51.75 drivers on an NVIDIA GeForce FX 5900 Ultra graphics card. Driver settings are shown on the graphs as well.

click her please for the rest:
http://www.amdmb.com/article-display.php?ArticleID=257


:cheers:

The reason he doesn't use a DX9 game is because they don't support it!

 

[DimitriPopov]

Originally posted by TrueWeltall
To make money.

Right here youre being very shortsighted. Nvidia is a coropration that has been fairly successful. They dont get there by being morons. They might make money for a few weeks but then when this information came out they would be fcked. No , they wouldnt do that , no way no how. So they are probably not 'cheating' and the fact that they could include something that makes benchmarks better shows how stupid benchmarks really are. If they dont reflect real in-game performance then they are completely useless.

 

[TrueWeltall]

Originally posted by DimitriPopov
If they dont reflect real in-game performance then they are completely useless.

Tell that to people who buy graphics card over benchmarks. Thats how it is today.

So they are probably not 'cheating'

So they never fixed with 3d mark? Yeah right.

They might make money for a few weeks but then when this information came out they would be fcked.

Thats how it is right now.

 

[DimitriPopov]

Originally posted by TrueWeltall




Thats how it is right now.

Ill say it again. They didnt get to their current place by being FRIGGIN MORONS. They know what will be profitable. They are not bumbling idiots.

 

[TrueWeltall]

Originally posted by DimitriPopov
Ill say it again. They didnt get to their current place by being FRIGGIN MORONS. They know what will be profitable. They are not bumbling idiots.

Haha

 

[shapeshifter]

Originally posted by DimitriPopov
Right here youre being very shortsighted. Nvidia is a coropration that has been fairly successful. They dont get there by being morons. They might make money for a few weeks but then when this information came out they would be fcked. No , they wouldnt do that , no way no how. So they are probably not 'cheating' and the fact that they could include something that makes benchmarks better shows how stupid benchmarks really are. If they dont reflect real in-game performance then they are completely useless.

you DO know it was proven that Nvidia was/has been cheating alot in the past right? not specific things to make the game go faster but Real no lite Just the real deal cheating, things like clipping planes being removed etc, what gabe is saying is "guess what? there still doing it"