nVidia, nVidia, nVidia..... [Archive]

micamica1217

11-12-03, 02:07 AM

I always save the best for last......Image Quality Differences!!!!!!!! (http://www.beyond3d.com/articles/3dmark03/340/index.php?p=6)

ok, very few peeps will notice anything showed on this page.
yet it is clear that the biggest differances can be clearly seen in GT3....hair, sword

but take a look at what is said by Derek Perez:

""NVIDIA asked for a conference call with us to go over some of the points, and in general some of the responses from Derek Perez were fairly frank and honest. It would seem that NVIDIA agrees with tome of the optimisation policy that Futuremark have put in place, but not all as they don't necessarily tally with the internal policies that NVIDIA have. From the gist of the conversation it seems that NVIDIA feels that it is their right to optimise specifically for applications as long as it doesn't interfere with image quality - which is fine if it is benefiting games as it will benefit the end user, however in this case the image quality can be argued as there are differences in the image produce with the optimisations that NVIDIA have used in the 52.16 drivers for the 3.3.0 patch. Derek went on to suggest that they may well end up chasing each patch release and re-optimising as Futuremark puts a patch that defeats previous detections!

At one (point)we asked Derek how this sat with the optimisations guidelines that were given to press by NVIDIA, specifically the guideline that suggests "An optimization must accelerate more than just a benchmark" To which Derek's reply was "But 3DMark03 is only a benchmark" -- it was suggested that this particular guideline should read "An optimization must accelerate more than just a benchmark unless the application is just a benchmark"! ""(article was dated November 12)

hmmmm....Nov. 3rd, nVidia spells out there optimasation policy on "editors day":
We will continue to do application specific performance optimizations and compatibility fixes. We met with Futuremark and discussed the optimizations they questioned. Some of the optimizations (such as intelligent color clears and clip planes) they viewed as too aggressive, so we have removed them from our drivers. We have new, stricter guidelines in place to insure that questionable driver optimizations do not get introduced in future driver builds including:

An optimization must produce the correct image
An optimization must accelerate more than just a benchmark
An optimization must not contain pre-computed state

I realy like what follows on the same page....

""With respect to NVIDIA's optimisation guidelines, questions should be asked exactly how stringent they are being applied, or whether they are just for the purposes of appeasing the press. One the face of it here, the optimisations applied here appear to break all three guidelines - the correct images do not appear to have been produced with the 52.16 drivers, pre-computed state in the form of shader replacements appear to have been used (if this were part of a generic optimisation then it should not have been defeated by simply altering the order of instructions), and hence these optimisations would be specific to this benchmark only. Given that we first heard talk of NVIDIA's optimisations policy during the Detonator 4x.xx release of drivers there would appear to have been ample time to have removed these types of optimisations by now.""

full read here. (http://www.beyond3d.com/articles/3dmark03/340/)

mica

Steven4563

11-12-03, 04:00 AM

im not an nVidia fan nor a ATi fan seeing ive had both. I just dont know y people cant leave it to rest and move on

unreal

11-12-03, 04:19 AM

lool

Cowboy X

11-12-03, 06:52 AM

I just posted this , I thought this thread was named something else and whne I couldn't find it I just posted this same info elsewhere :( . Anyway it just goes to show that NV doesn't care about guidlines ,rules or what gamers and the industry think . Until fully checked , no benchmark involving them can be trusted especially if it is popular . Just look at Mr.Perez's comments ! Is anyone surprised by Valve's statements earlier in the year , Mr. Perez has basically said that they intend to continue cheating with each new patch/driver !

Steven4563

11-12-03, 08:27 AM

then if they are always gonna cheat y dont the companies that make the games and benchmarks take them to court ?

harryinny3

11-12-03, 08:29 AM

I dont know what res and all those Comparison IQ pics were at, But my 9600pro with the 3.9 drivers, My IQ in the benchmark are atleast 75% better and Alot sharper looking.

As for the nvidia Socalled Opti's. Im glad more and more i didnt buy that 5600ultra FX i was gonna but instead of this card. Yikes.

Harry

JigPu

11-12-03, 10:00 AM

Originally posted by harryinny3
I dont know what res and all those Comparison IQ pics were at, But my 9600pro with the 3.9 drivers, My IQ in the benchmark are atleast 75% better and Alot sharper looking.
I'm wondering if there ARE any IQ differences inherant to the patch... The article says that there shouldn't be, but they never mentioned running a test on a reference board which dosen't cheat (possibly ATi, though to be truly ubiased, you'd need somebody else). This means that IQ differences could have popped up between the patches, making the FX look anywhere from a lot worse than it is, to just a bit worse than it is.

I'm willing to bet that at least 80-90% of the difference shown isn't a result of the patch changing things (and thus a result of the FX's cheats being thwarted), but I much prefer definate proof on an issue (specifically, this patch issue) to near-definate proof.

Anybody here willing (or know how) to take screenies from 03 at the frames they used?
JigPu

harryinny3

11-12-03, 10:16 AM

I use to have a program that took snapshots of anything i ran. I used it to Report Abusers in my yahoo pool league when i was a mod. Ill look it up and get back to you. It just might work.

Harry

micamica1217

11-12-03, 02:54 PM

jigpu,

you'll need the pro vertion of 3dmark03.....so you may take a snapshot of a perticular frame.
(say you want to show frame #125 with two different cards)

mica

Wassup Doc

11-12-03, 03:20 PM

good to see the futuremark finally decide to not be nvidia's bitch.

micamica1217

11-12-03, 03:33 PM

Originally posted by Steven4563
im not an nVidia fan nor a ATi fan seeing ive had both. I just dont know y people cant leave it to rest and move on

well lets see, I'm not going to ignore this because I love futuremark, or because I love ATI.
it's not because I feal sorry for the developers who spend weekends away from thier family in order to put out a game on time.
nor do I realy care if such said devleloper is now upset that a consumer with a nVidia card may not get to "see" what is intended for them to "see".

it is because, I am a consumer, and I feal that WE consumers are being ripped off by nVidia with there cheating drivers that show time and again the poor IQ(lack of images in some cases) of the FX cards.
maybe you don't care if your suposed to see a shadow under a ship traveling across a field. but I want to see that shadow, rock, or anything that I'm suposed to see.
for example:
could you just imagene, spending $300-$400 on a new vid card and then spending $50.00 on Halo.....
then as you are playing it you find out that if you got a different card you would have seen pipes and wires on the ship(that you are unable to view at this time). you'll have more fog in the game.
and finaly if you are outside, durring the day, you will have more lighting on objects shuch as the rocks by the water fall.
I can't speak for you or others, but I know that I would feal ripped off.

I can't realy understad how nVidia can believe that we the consumer would be happy with such a loss of IQ.

mica

Cowboy X

11-12-03, 08:13 PM

Originally posted by Steven4563
then if they are always gonna cheat y dont the companies that make the games and benchmarks take them to court ?

If Nvidia or any other Vcard maker breaks the license agreement of 3dMark 2003 they can be sued . This agreement includes provisions that allow you to publish results in which you have conform to Futuremark's rules . What Nvidia is doing is cheating but websites and reviewers publish the results and Nvidia is in the clear .

What game developers and benchmark makers have to do is to set strong rules and enforce them , and when the rules are broken have no qualms about telling all . When hardware makers see top games and software having their hardware listed as substandard or unsupported or not recommended ( with reasons given ) then things will begin to change . If Nv or anyone is silly enough to then sue the software maker for hurting their product image they will lose .

The other ways a wayward hardware maker can be sued are :

1/ Consumers sue due to false advertising and fraud .
2/ A software developer can actually prove that the maker's false claims and or cheating in their software hurt their software's sales .
3/ The hardware maker erroneously blames the software maker in public for its hardware's flaws .

Mike360000

11-12-03, 08:28 PM

I don't doubt the ability of nvidia to cheat but take a looky at this article concerning Gainward, from Xbit Labs. Gainward claims that the patch 340 disables the whole compiler for the fx gpu thus crippling the v card. They claim the fx v cards will run all game's codes equally, including 3DMark2k3.

Linky:
http://www.xbitlabs.com/news/video/display/20031112031947.html

Mike

HardwareJedi

11-12-03, 10:13 PM

A Response from Futuremark concerning Gainwards allegations...

"Futuremark has made an official comment to Gainward his accusations:
The accusation is totally wrong because what it suggests is not even feasible technically. 3DMark03 does not talk to graphics driver, it talks to the DirectX API, which then talks to the driver. Thus, it is impossible for the application to disable GPU compiler.

The only change in build 340 is the order of some instructions in the shaders or the registers they use. This means that new shaders are mathematically equivalent with previous shaders. A GPU compiler should process the old and the new shader code basically with the same performance. Of course, if there are application specific optimizations in the driver that depend on identifying a shader or parts of it, then you might see performance differences because these optimizations will not work if the driver is not able to detect the shader.

Let's also repeat that 3DMark specific driver optimizations are forbidden in our run rules because they invalidate the performance measurement and the resulting score is not comparable to other hardware.

Thus, the right conclusion is that the new version of 3DMark03 is now very suitable for objective performance measurement between different hardware."

It's (The Gainward Claim) Just more Nvidia FuD campaign material to discredit anybody that shines a light on their subterfuge'

Cowboy X

11-12-03, 10:15 PM

That is an interesting claim from a manufacturer , but I'd rather hear it from Nvidia . More importantly the driver compiler in the 52.16 does automatic shader replacement ! This invalidates synthetic benchmarks immediately since the card does something different from the rest of the cards and thus can't be compared anymore as apples to apples . What is further the problem is that these shader replacements are also probably part of the reason that NV's cards show different images from the reference . I'm therefore not surprised that Futuremark blocked them .

Cowboy X

11-12-03, 10:22 PM

Looking at what Futuremark has said in reply , I think Gainward meant to say that the driver compiler was blocked , at least I hope that is what they meant .

micamica1217

11-12-03, 11:09 PM

Originally posted by Cowboy X
Looking at what Futuremark has said in reply , I think Gainward meant to say that the driver compiler was blocked , at least I hope that is what they meant .

that is what they meant to say....and I call bullcrap.

blocking the compiler in any way should see a performance drop of equal value to the vertex shader test in PS2.0 test....
as well as higher performance drops in all tests.
plus, there would not be any image changes like the ones listed in one of my above replys.

maybe I'm way wrong on this.....
but I'll keep calling bullcrap, bullcrap, bullcrap, with each driver that shows cheating from nVidia.
I'll also believe the boys from beyond3d over gainward/nVidia any day of the week.

you know I'll find out more, and list it here.
(even if it helps nVidia in any way)

mica

Cowboy X

11-12-03, 11:28 PM

I don't disagree with you but Gainward saying the driver compiler is alot less silly an argument than saying the GPU compiler .

micamica1217

11-12-03, 11:30 PM

Originally posted by Cowboy X
I don't disagree with you but saying the driver compiler is alot less silly an argument than saying the GPU compiler .

I understand compleatly what your saying.

mica

looking closer, nVidia confirms that it is the "GPU compiler".....

""An official from NVIDIA Corporation confirmed Mr. Tismer’s accusation that “patch 340 disables the GPU compiler. The compiler has to run on the CPU instead resulting in code harder to digest and taking away 20% of the performance.” “Yes, that is actually the case with the new patch 340 that Futuremark posted,” said an NVIDIA spokesperson on Wednesday.

also added...

""“Few weeks ago we released our 52.16 driver that includes our brand new unified compiler technology. With the new patch the benchmark, our unified compiler gets not used by the app so it goes to CPU and we are definitely slower,” Luciano Alibrandi, NVIDIA’s European Product PR Manager, added.""

OK boys and girls, riddle me this....
how does a benchmark not use the "GPU/unifide" compiler and only a few tests go down in the benchmarking scores????
if the CPU must handle the code, shouldn't it have to do it with ALL the tests???

only a few tests show a performance loss....
almost everything else, is well within the margin of error.

only game tests "4" and "vertex shader" have been hit the hardest....GT2 and GT3 was hit some, yet "Pixel Shader 2.0" test was not hit at all.
it would seem to me that at least PS2.0 test would be hit at least as much as GT2 and GT3, if not as much as the vertex shader test.

*mica wonders why nVidia keeps opening there mouths???
don't they realise that they are just sticking a body part into it?*

Matthias99

11-13-03, 11:14 AM

So... basically what NVIDIA is saying is that if they've "optimized" for a piece of software, and the software's developers make even the slightest change to their software, NVIDIA's optimizations will be broken and no longer work?

If their compiler technology relies on application detection to acheive any sort of decent performance gain (and isn't getting away from app-specific optimization the whole point of the compiler?!), then it's basically useless. You'd need a new driver release every time they tweak anything in Doom3 or HL2.

sautegod

11-13-03, 11:24 AM

so they have released a patch, big deal
so now nvidia will release a new driver that fixes that.

Cowboy X

11-13-03, 12:13 PM

Well micamica1217 , I guess Nvidia's lame and silly excuses get even more silly , they should have kept to their old fallback position " It's a bug "

Mike360000

11-13-03, 12:54 PM

Well it's their problems not mine......
All I do is play a few games, surf, a little video editing.
FutureMark and nvidia can slug things out among themselves.
But I like my 5900 fine, and it doesn't effect my decisions at this time.

Mike

Falcon-K

11-13-03, 01:15 PM

Originally posted by Mike360000
Well it's their problems not mine......
All I do is play a few games, surf, a little video editing.
FutureMark and nvidia can slug things out among themselves.
But I like my 5900 fine, and it doesn't effect my decisions at this time.

Mike

if you play games then it is your problem, nvidia constantly makes optimizations that make your gaming experience less, they remove stuff that helps the gamer to immurse themselves in the game, its the whole point in spending a ton on a videocard is so you can see everything and have it run smoothly. so in a way it is your problem

Mike360000

11-13-03, 01:35 PM

"if you play games then it is your problem, nvidia constantly makes optimizations that make your gaming experience less, they remove stuff that helps the gamer to immurse themselves in the game, its the whole point in spending a ton on a videocard is so you can see everything and have it run smoothly. so in a way it is your problem"

It is all subjective and arguementative.
Plus the whole idea of knowing every last small detail on what ANY video card maker does with their hardware and coding is not possible. OTOH what is possible is seeing and knowing what one likes. And that is a totally subjective matter.

Mike

Hurk

11-13-03, 02:03 PM

I'm a gamer, and if I knew, or even thought I was missing out on something, I'd feel ripped off. If I though the IQ is be better on the other brand of card, I'd feel ripped off. If my new card that cost more than other cards, doesn't perform aswell as the other card, I'd feel ripped off. I don't care if they took out one bar, or glare or whatever, the point is that they ARE doing that, and the other company is not.

Th0r

11-13-03, 02:26 PM

Mmmmm.... I wounder what will Nvidia will pull next....

micamica1217

11-13-03, 04:53 PM

Originally posted by Matthias99
So... basically what NVIDIA is saying is that if they've "optimized" for a piece of software, and the software's developers make even the slightest change to their software, NVIDIA's optimizations will be broken and no longer work?

If their compiler technology relies on application detection to acheive any sort of decent performance gain (and isn't getting away from app-specific optimization the whole point of the compiler?!), then it's basically useless. You'd need a new driver release every time they tweak anything in Doom3 or HL2.

nVidia's compiler technology doesn't rely on application detection.
but there are application specific optimizations in thier drivers.

in the case of 3dmark03, nVidia has replaced some of the shaders that were in the app. with thier own made up ones.
thus getting a higher score, yet not getting the same image as in the referance image.
in some cases, there are missing objects...such as there are missing stars.
or, the IQ is lowered on some of the objects, such as the floor on the ship.
both cases are hardly noticable to most people.
but then you have the GT3....
missing hair and lower detail on that said hair is plain to see by all.

application specific optimizations will not be broken with a patch to update a game...unless that is the intent of the patch.

------------

btw, it seems that nVidia is more conserned right now with optimizing thier drivers on a more game spicific rought, then say trying to have all DX or OpenGL run faster.
giving some games great IQ with good speed increases, while others get poor IQ in order to get a great speed increases.

the problem I have is that nVidia is no longer giving me a choice of how good the IQ will be when I run a game.
while I could always lower the AA, AF, or texture quality to suit my needs with ATI.

mica

Th0r

11-13-03, 05:19 PM

Well, I have had only 2 cards with Nvidia....a GeForce 4MX 420MX and the 440MX....

I have had a 3dfx voodoo 3 2000 AGP.... A Rage 128 Pro, RVE,R8500...... u have loved these cards...the Nvidia were a bad choice ...

Matthias99

11-13-03, 06:51 PM

Originally posted by micamica1217

application specific optimizations will not be broken with a patch to update a game...unless that is the intent of the patch.

If NVIDIA is relying on detecting *exact* shaders for replacement (which it appears they are doing, since tiny changes to the 3DMark03 shaders have screwed it up), then *any* changes to the shader code in a DX9 game will "turn off" some or all of NVIDIA's replacement shaders, thus necessitating a driver rerelease (a small one, but still -- wasn't NVIDIA saying just a few weeks ago that they wanted to move to 1 release per YEAR?!) to reactivate them.

micamica1217

11-13-03, 10:29 PM

Originally posted by Matthias99

If NVIDIA is relying on detecting *exact* shaders for replacement (which it appears they are doing, since tiny changes to the 3DMark03 shaders have screwed it up), then *any* changes to the shader code in a DX9 game will "turn off" some or all of NVIDIA's replacement shaders, thus necessitating a driver rerelease (a small one, but still -- wasn't NVIDIA saying just a few weeks ago that they wanted to move to 1 release per YEAR?!) to reactivate them.

what I'm getting at is that unlike in 3dmark03, niVidia is not normaly replacing shaders in order to get an extra performance boost in games.
what they are doing is running ALL DX games with "bri-liner" filtering....yes, it's not just UT2k3.
AF is almost always reduced in all DX games unless they pump it up a notch to were it should be anyway.(so you turn it on to 4xAF, yet it realy gives you the quality of 2xAF and a performance boost....at least that's the IQ that I'm seeing in most, but not all games.)
AA IQ is on a per game bases as to weather it gives you the true AA that you preset or not.
for example, UT2K3 it works great.....yet a few others like MS flight sim 2k4 it works, but maybe not at the true setting(set it at 6X and get 4X in game).

I'm not saying that nVidia is not replacing textures or shaders...
yet at this time, I see no evadence that sugjests this for games that are out.
what I will state is that nVidia is also reducing the amount of lighting in some games.(halo and aquanox/aquamark are big offences)

mica

micamica1217

11-14-03, 12:10 AM

as seen from HardOCP........

Derek Perez with NVIDIA's PR Dept replied:
""This latest patch from Futuremark is yet another revision of 3DMark03 specifically designed to defeat our Unified Compilier Technology, which evaluates shaders and in some cases substitutes hand tuned shaders, but increasingly simply applies the run-time compiler to generate optimal code. With the 52.16 drivers and the new patch, our perf drops 15%.

Clearly our compiler has gotten much better, as image quality remains exactly the same, the only thing that happens is a 10-15% drop in performance.

We're not sure why anyone would want to reduce their performance by 10-15% for the same image quality, but apparently Futuremark feels that is something relevant.

What we expect will happen is that we'll be forced to expend more engineering effort to update our compiler's fingerprinter to be more intelligent, specifically to make it intelligent in its ability to optimize code even when application developers are trying to specifically defeat compilation and optimal code generation.

This is yet another example of how 3DMark03 doesn't behave like a game - as a game developer would never specifically try to make their application run poorly or disable optimizations that produce the correct image while delivering better performance.""

I like the way he's saying "same image quality" over and over again.:rolleyes:

I'm not saying that there are tons of abnormal IQ problems before the relise of the new build....
yet there are times when you can see that there is a problem or two, and there is no need for any "image test" for me to know.
don't the PR people realise that we are not dumber than them???

mica

Mike360000

11-14-03, 12:40 AM

"don't the PR people realise that we are not dumber than them???"

mica that was a good'un!
That is like saying don't drs. realise that we are not dumber than them??? Say especially when it comes to medicine?
And the same could be applied to lawyers, mechanics, teachers, or anyone who likes to think all their hard spent cash for some degree somehow makes them more intelligent than others.

Don't get me wrong, cuz I agree with you, but people really do have this mentality that only they are qualified to speak concerning their field and anyone who isn't in that field isn't qualified to say anything contrary to what they say. At least not say it with any credibility.

I will add though, in giving nvidia the benefit of a doubt. That it would seem possible given nvidias coding/driver writing abilities, that it may be possible to do some or most of what D. Perez said. While I can't say it for a fact, nor could hardly any person who isn't in the highest of positions to know, it would still seem possible. I think our answers lies in finding out exactly what their so called "optimized" coding really does.......

Mike

micamica1217

11-14-03, 01:06 AM

as posted from beyond3d.....

Update 13/11/2003

Futuremark have told us that some the the smoke from the door in GT2 and the hair and fire shaders in GT3 may not (be) exactly the same from one run to the other. These should be discounted as image differences.

(see I'm trying to be fair to nVidia)
yet I'll resurve my thoughts for a later time....

mica

Matthias99

11-14-03, 01:10 AM

Originally posted by micamica1217

what I'm getting at is that unlike in 3dmark03, niVidia is not normaly replacing shaders in order to get an extra performance boost in games.

They can't do it right now because DX8.1 games (every game you mentioned besides Halo, and that only uses a little DX9) *don't use pixel and vertex shaders*. If they're replacing shaders with 3DMark03, I find it likely they'll try to do the same sort of thing with HL2 and Doom3. But how many games can they optimize like this? Writing custom shaders by hand -- especially for longer and more complex effects -- is very time-consuming.

Mike360000

11-14-03, 01:22 AM

I read somewhere, that nvidia wanted game makers to use their complier and recompile games so as to include their specialized coding. Supposedly it work seperately from the standards for DX9. However I also read and posted that Microsoft will most likely do a update for DX9 to DX9.1 and that update would include coding more favorable to nvidia.

I find it odd that nvidia's own coding, which includes 32 bit precision, and which supposedly should be better than the official DX9 standards of 24 bit precision, that nvidia is having IQ problems dropping some textures, as well as those textures changing from test to test. I know the 32 bit precision is not useful at this moment but still I find the whole current situation quiet odd unless nvidia is really fudging. OTOH maybe their optimizations are so complex and as was said so time consuming that all kinds of quirks are taking place?

Mike

Falcon-K

11-14-03, 02:13 AM

Originally posted by Mike360000
I find it odd that nvidia's own coding, which includes 32 bit precision, and which supposedly should be better than the official DX9 standards of 24 bit precision, that nvidia is having IQ problems dropping some textures, as well as those textures changing from test to test. I know the 32 bit precision is not useful at this moment but still I find the whole current situation quiet odd unless nvidia is really fudging. OTOH maybe their optimizations are so complex and as was said so time consuming that all kinds of quirks are taking place?

Mike

i think the reason that they lose IQ is because most times nvidia cards use 16 bit percision which is not as good as the dx9 standard 24 bit percision and when they run in 32 bit percision the cards come to a hault and just perform horrible, so nvidia is forced to either lose IQ or lose performance both of which are not very good options so nvidia came up with the third option they will drop some of the lighting, details, fog whatever it maybe so that everything appears to be a high level of IQ and it runs well. But as we all know people notice when things are missing, its just a matter of time and then they call it a bug. If im wrong it wont be the first time.

Evnas

11-14-03, 02:26 AM

Originally posted by Mike360000
"Don't get me wrong, cuz I agree with you, but people really do have this mentality that only they are qualified to speak concerning their field and anyone who isn't in that field isn't qualified to say anything contrary to what they say. At least not say it with any credibility.

The funny part about that, is the PR Dept knows NOTHING about the complex workings of the card nor the drivers. They know what they are told to say, and make the company look good. Basically an over payed Sales Associate, heh

micamica1217

11-14-03, 04:04 AM

Originally posted by Falcon-K

i think the reason that they lose IQ is because most times nvidia cards use 16 bit percision which is not as good as the dx9 standard 24 bit percision and when they run in 32 bit percision the cards come to a hault and just perform horrible, so nvidia is forced to either lose IQ or lose performance both of which are not very good options so nvidia came up with the third option they will drop some of the lighting, details, fog whatever it maybe so that everything appears to be a high level of IQ and it runs well. But as we all know people notice when things are missing, its just a matter of time and then they call it a bug. If im wrong it wont be the first time.

I don't think your too far off on this one....

the first thing that came to my mind when I saw the edit at beyond3d ""Futuremark have told us that some the the smoke from the door in GT2 and the hair and fire shaders in GT3 may not (be) exactly the same from one run to the other. These should be discounted as image differences."" ...
was how could a benchmark with prescripted video, display different images on two different runs????
isn't the benchmark like a DVD movie???
run it the second time, and all the parts are the same, right???

then it hit me....
nVidia is running with 16bit VS2.0.....maybe.
that would explain the loss in detail on the floor, and the loss of some of the stars by default.
going with 32bit VS2.0 would also now give you more hair and more detailed hair as well as lower performance (maybe 16bit was used here as well).
it would also be understandable that the PS2.0 test would not go down.
as for the vertex shader test...maybe it too was using 16bit.

so why would futuremark back nVidia up in any way???
didn't futuremark already do this once before?

I would realy like to ask a few more questions about this before I make any false claims.

mica

micamica1217

11-14-03, 04:28 AM

Originally posted by Matthias99

They can't do it right now because DX8.1 games (every game you mentioned besides Halo, and that only uses a little DX9) *don't use pixel and vertex shaders*.

Oh yes they do. it may be just PS1.4 or less, but they all use some sort of vertex and/or pixel shaders.

If they're replacing shaders with 3DMark03, I find it likely they'll try to do the same sort of thing with HL2 and Doom3.

DOOM3 should also be more like a DX8.0 game that only uses a little DX9...like halo.
yet, I have to disagree that they most likely will not replace any shaders in DOOM3 as well as HL2...yet who knows?

But how many games can they optimize like this? Writing custom shaders by hand -- especially for longer and more complex effects -- is very time-consuming.

just my two cents

mica

HardwareJedi

11-14-03, 06:28 AM

Mike: "OTOH maybe their optimizations are so complex and as was said so time consuming that all kinds of quirks are taking place?"

Which to me means that the FX is a shoddy piece of work, requiring extra effort on the part of the Game developer to get any kind of decent performance out of an FX card. Nv let the Marketing Dept. blaze the trail for the FX with claims of Cinematic Computing, and FP32 precision, and Unlimited Shader instruction length etc etc etc, and then the Silicon shows up and guess what? None of the Marketing check box claims will run fast enough to be useful.
Want some insight into why the FX sucks so bad, read this...

http://www.notforidiots.com/ULE1.php

Mica, as far as Doom III is concerned, Thanks to J. Carmack, Doom III should run just fine on Nv Hardware. It has a specific code path designed just for the FX. So I agree that they may not need to replace shaders in Doom III.

PreservedSwine

11-14-03, 08:20 AM

Originally posted by Mike360000
"don't the PR people realise that we are not dumber than them???"

mica that was a good'un!
That is like saying don't drs. realise that we are not dumber than them??? Say especially when it comes to medicine?

Mike

WRONG

More like car salesman or advertiser knowing more about a product than an enthusiast.

You do know what PR people do, right? If you don't know by bow, you're a PR guy's dream....

HardwareJedi

11-14-03, 08:22 AM

News Flash!!! Nvidia says Compiler NOT disabled by 3dMark03. Gainward Official full of **** (so to speak)

http://www.xbitlabs.com/news/video/display/20031114041519.html

excerpt:
"Luciano Alibrandi, European Product PR Manager for NVIDIA Corporation, has made a correction in regards previous information about NVIDIA’s Unified Compiler and 3DMark03 benchmark after getting into details with the company’s engineers. Apparently, the statement claiming that NVIDIA’s Unified Complier deployed to optimize pixel shader performance is disabled by the new version of 3DMark03 is not fully correct.

“I would like to inform you that a part of my response was not accurate. I stated that the compiler gets disabled, by 3DMark and that is in fact not true,” he said.

So, after all NVIDIA denied the problems between the Unified Compiler technology and the latest version of popular 3DMark03 benchmark. As a result, we may now conclude that the accusations in Futuremark direction from Hans-Wolfram Tismer, a Managing Director for Gainward Europe GmbH were not correct at all. "

Edit: Ok so the Compiler is working ok. Then just what did cause the drop in performance??? Hmmm??? Gee couldn't be Bogus shader replacement being defeated by the patch could it? Naw! Nv wouldn't do that would they? :rolleyes:

PreservedSwine

11-14-03, 08:23 AM

Originally posted by HardwareJedi
Want some insight into why the FX sucks so bad, read this...

http://www.notforidiots.com/ULE1.php

What a great read!

I didn't want to let that link slip by undetected, that's some GREAT writing by someone who knows their stuff

PreservedSwine

11-14-03, 08:30 AM

Originally posted by Mike360000
"if you play games then it is your problem, nvidia constantly makes optimizations that make your gaming experience less, they remove stuff that helps the gamer to immurse themselves in the game, its the whole point in spending a ton on a videocard is so you can see everything and have it run smoothly. so in a way it is your problem"

It is all subjective and arguementative.
Plus the whole idea of knowing every last small detail on what ANY video card maker does with their hardware and coding is not possible. OTOH what is possible is seeing and knowing what one likes. And that is a totally subjective matter.

Mike

"A new PC game gets released about once a day; about one-third of these are games that really push 3D graphics. Only a tiny percentage of these will receive the dubious "optimizations" that have been directed at previous versions of 3DMark03. Gamers don't want to be locked into these. They don't want to be surprised by poor frame rates when they buy a game outside the top 10. Gamers need uncompromised benchmarks that give them a true picture of performance, so they can find a card that performs for all games, not just the half-dozen for which the drivers are faking it."

Not so subjective, Mike

PreservedSwine

11-14-03, 08:41 AM

Originally posted by Mike360000
I read somewhere, that nvidia wanted game makers to use their complier and recompile games so as to include their specialized coding. Supposedly it work seperately from the standards for DX9. However I also read and posted that Microsoft will most likely do a update for DX9 to DX9.1 and that update would include coding more favorable to nvidia.

I find it odd that nvidia's own coding, which includes 32 bit precision, and which supposedly should be better than the official DX9 standards of 24 bit precision, that nvidia is having IQ problems dropping some textures, as well as those textures changing from test to test. I know the 32 bit precision is not useful at this moment but still I find the whole current situation quiet odd unless nvidia is really fudging. OTOH maybe their optimizations are so complex and as was said so time consuming that all kinds of quirks are taking place?

Mike

Another take, re: compilers

It's been claimed that Futuremark's changes have disabled compilers. This is complete nonsense. ATI has had a compiler since CATALYST 3.6 and it didn't have any problems with Futuremark's changes. Shader replacement and compilers are completely different operations.

ATI has had a compiler since CATALYST 3.6. We didn't have any problems with Futuremark's changes. The new build of 3DMark03 gives an honest picture of the relative DX9 game performance of graphics cards. It accurately reflects what gamers will see with titles such as Half-Life 2, and what they already see with today's DX9 games, such as Tomb Raider: Angel of Darkness.

Secondly, it is disingenuous to claim that Shader replacement better reflects the performance in games. Only a tiny fraction of games get the attention that 3DMark03 has had from our competitors. It requires too much programming time. Over 300 PC games are launched a year, and 100 of these will really tax the graphics hardware. Maybe a half-dozen - the ones most used as benchmarks - will receive the gentle caress of the driver engineer. An honest run of 3DMark03 will give a true indication of performance for the overwhelming majority of DirectX 9 games. Gamers need to be able to play any game they want; they don't want to be locked into the six that have had all their shaders replaced.

Even assuming that you somehow found the resources to replace all the shaders in every game, it's still not a practical solution. Shader replacement is a massive step back for reliability and game compatibility. Every year you'll be writing thousands of Shader programs that each have to be checked for image quality, taken through QA and supported. And changed whenever the developer issues a patch. Treating every game as a special case is a huge stability issue. Developers have come out against Shader replacement. John Carmack is on record as saying "Rewriting shaders behind an application's back in a way that changes the output under non-controlled circumstances is absolutely, positively wrong and indefensible." The opinions of Gabe Newell, Valve Software's CEO, on Shader replacement are well-known. Developers hate it. What if they release a new level, the gamer downloads it and performance sucks? The hardware vendor isn't going to get any grief, because all the user sees is the old levels working fine and the new one running like molasses in January. The problem's obviously with the game, right? Developers are worried about the support nightmare this approach will generate and the damage to their own brand when they get blamed.”

Oh, and Mike, nVidia doesn't use FP32, although the hardware is capable. It's too slow at FP32, so they must use FP16, FP12, and even (gasp) FX12

Matthias99

11-14-03, 11:00 AM

Originally posted by micamica1217

just my two cents

mica

Sorry, I should have been more explicit that I meant PS2.0 (DX9-spec) pixel shaders, which is where NVIDIA's performance tanks. :) They do OK at 1.0/1.2/1.4 shaders natively, mostly because many of them are very very short. :)

micamica1217

11-14-03, 11:27 AM

Originally posted by HardwareJedi
News Flash!!! Nvidia says Compiler NOT disabled by 3dMark03. Gainward Official full of **** (so to speak)

http://www.xbitlabs.com/news/video/display/20031114041519.html

excerpt:
"Luciano Alibrandi, European Product PR Manager for NVIDIA Corporation, has made a correction in regards previous information about NVIDIA’s Unified Compiler and 3DMark03 benchmark after getting into details with the company’s engineers. Apparently, the statement claiming that NVIDIA’s Unified Complier deployed to optimize pixel shader performance is disabled by the new version of 3DMark03 is not fully correct.

“I would like to inform you that a part of my response was not accurate. I stated that the compiler gets disabled, by 3DMark and that is in fact not true,” he said.

he also stated ""Since the complier is still active with the new version of 3DMark03 there is currently no explanations for performance drops of certain GeForce FX parts in the latest build 340 of the famous 3DMark03.""

Edit: Ok so the Compiler is working ok. Then just what did cause the drop in performance??? Hmmm??? Gee couldn't be Bogus shader replacement being defeated by the patch could it? Naw! Nv wouldn't do that would they? :rolleyes:

maybe your right, and nVidia did some shader replacement....
but unlike what Luciano Alibrandi said, there is an explanation as to why there is a perf drop.

I still feal for many reasons, that my thoughts on nVidia using only 16bit VS2.0 (or a lower precision of VS) with posably an app detection could be the real anser.

what are your thoughts on this PreservedSwine?

mica

micamica1217

11-20-03, 08:38 AM

update:

well it seems that some things are just more unbelievable....
I wasn't far off with my belief that maybe 16bit VS2.0 is used for 3dmark03.

first, I found out that with all FX cards, they are unable to run in any DX game/benchmark with VS2.0, higher then 16bit FP

also from beyond3d:

I've done some more tests around the 3Dmark PS2.0 test.

FX5950 52.16 : 50,8
FX5950 45.23 : 27,1
9800XT 3.9 : 52,4

This PS2.0 test uses 2 big FP24/FP32 2.0 shaders. I've used them in another application.

Shader 1 (MPix/s):
FX5950 52.16 FP32 : 71,1
FX5950 52.16 FP16 : 99,4
FX5950 45.23 FP32 : 74,1
FX5950 45.23 FP16 : 98,8
9800XT 3.9 FP24 : 112,1

Shader 2 (MPix/s):
FX5950 52.16 FP32 : 49,1
FX5950 52.16 FP16 : 61,7
FX5950 45.23 FP32 : 34,9
FX5950 45.23 FP16 : 61,3
9800XT 3.9 FP24 : 71,3
----------

well since it seems that in most cases the FX cards get a boost from the new drivers in PS2.0 32bit....
it seems that they may be using 16bit for the PS2.0 test since there was no performance boost with or with out the new build.

next is a rightmark3d test: taken from a forum who a few members tested this progie.

""Interesting results Hans and mildewing - looks like the vertex throughputs of the R300 is virtually identical, regardless what method is used. Compare those to a 5900U (52.16 drivers) and a 5600U rev1 (45.23 drivers):( this is rightmark3d-benchmark)

Fixed function
9800P = 41,537,500 poly/s
9700P = 36,336,848 poly/s
5900U = 59,287,784 poly/s
5600U = 14,745,670 poly/s

Vertex Shader 1.1
9800P = 38,791,528 poly/s
9700P = 33,203,558 poly/s
5900U = 39,571,376 poly/s
5600 = 9,067,759 poly/s

Vertex Shader 2.0
9800P = 38,708,316 poly/s
9700P = 33,126,878 poly/s
5900U = 14,860,529 poly/s
5600U = 5,783,667 poly/s

9800P = 380/680MHz (380Mverts/s - 4 vertex "units")
9700P = 325/620MHz (325Mverts/s - 4 vertex "units")
5900U = 400/850MHz (300Mverts/s - 3 vertex "units")
5600U = 350/700MHz (88Mverts/s - 2 vertex "units") ""

this last test was just to show what the true performance loss is like with the FX cards....note the almost no loss from the ATI cards.

I may have more info soon.......

mica