Why is NVIDIA against 3DMark03? [Archive]

Nutty

02-16-03, 08:00 PM

I dont think their current line lacks the technology to be good at DX9.

I maybe reading between the lines here, but it seemed the 9700 ran the DX9 scene's PS 1.4 shaders as PS1.4. It seems the GF-FX had some problem with this, and so promoted all the shaders to PS2.0, which is why it seems to run like a dog compared to the 9700 in the pixel shader category.

If some companies get some decent quiet cooling solutions on the gf-fx, then I'll probably consider getting one. I think given a few months for drivers to mature, we'll see it pull ahead from the 9700 alot more.

Sazar

02-16-03, 08:59 PM

Originally posted by Nutty
I dont think their current line lacks the technology to be good at DX9.

I maybe reading between the lines here, but it seemed the 9700 ran the DX9 scene's PS 1.4 shaders as PS1.4. It seems the GF-FX had some problem with this, and so promoted all the shaders to PS2.0, which is why it seems to run like a dog compared to the 9700 in the pixel shader category.

If some companies get some decent quiet cooling solutions on the gf-fx, then I'll probably consider getting one. I think given a few months for drivers to mature, we'll see it pull ahead from the 9700 alot more.

what are you on about?

the fx does what it is supposed to... its ps 2.0 does not seem as efficient as the 9700pro's for some reason @ the moment.. either because it is not or because of driver problems... dunno...

what problem exists is the dx8 gf4ti cards running the dx8 games with ps 1.3 instead of 1.4 hence using extra renderng pases and therefore garnering a lower score...

concerning comments about 3dmark03... READ THE WHITE PAPER BEFORE MAKING RETARDED COMMENTS

the original white paper perhaps...

maybe is nvidia or hardocp or toms had read the white paper they would not have posted half of what they did... as is I am hoping that the rumor mongers read futuremarks newly released .pdf answering all allegations about their product levied by nvidia...

John Reynolds

02-16-03, 09:03 PM

Originally posted by Smokey
It also DOES not reflect game performance. The 8500 was faster in 3dmark than my gf3, yet in games it wasnt :rolleyes:

Its good for system tweaks, to see what gives better performance from your machine.

Oh ... and what Nutty said :p

The 8500 scores well because of PS1.4 support. 3DMark, as someone over at B3D's board said, is supposed to represent more than just a card's performance, but also its "goodness" (technology, features, etc.). Make of that what you will.

Myrmecophagavir

02-16-03, 09:32 PM

Originally posted by John Reynolds
The 8500 scores well because of PS1.4 support. 3DMark, as someone over at B3D's board said, is supposed to represent more than just a card's performance, but also its "goodness" (technology, features, etc.). Make of that what you will.Bingo. I wish they wouldn't advertise it as "the gamer's benchmark", because maybe it's true that games don't use the same techniques (ie. PS 1.4). But if you view 3DMark as a test of what a card is capable of rather than trying to emulate gaming performance, it works better. 8500 can utilise PS 1.4 to get things done more efficiently than GF3, so why shouldn't it "win"?

It's like saying "The 8500 wins in app A, but the GF3 wins in app B, therefore app A is a bad benchmark for the card's performance". It's only that app A's programmers made an effort to take advantage of the 8500's extra features. If game developers would take advantage of it more widely then it would win in more tests!

ZoinKs!

02-16-03, 09:46 PM

Originally posted by Smokey
It also DOES not reflect game performance... The new 3dmark is not made to measure performance in current games. It is made to predict performance in games which will be released 1-2 years from now.

A graphics card producing a high 3dmark03 score will be better in future games than a card with a low 3dmark03 score.

If you want to know performance in current games, then look at framerates in those games. If you're shopping for a new card and want one that will last, then look at 3dmark03 results.

Skynet

02-16-03, 11:44 PM

Honestly, I've seen better stuff in a 64k demo.
Uh huh. So I guess we don't really need DX9 at all just code everything in a 64k boundry and that's all it takes.

You know, why doesn't everyone read FutureMark's white paper on 3DMark03 it explains a lot and answers many questions and concerns people are having.

AND GET IT THROUGH YOUR HEADS that it is for FUTURE GAMES not what is being played now. It's not that hard to understand. The reason the 8500 does better than expected on 3DMark03 is because it has more advanced shaders than a GeForce3, it's direct competitor at the time. That is not a rant or an opinion it is a FACT, why is that so hard to understand? The 8500 is most definitely slower and pretty much every game out there compared to a GeForce4 and most GeForce 3's, but it just so happens to have a better shader, no big deal. I'm not saying it will do well on a DX9 game because it won't it does not have the memory bandwidth to start with....

Face it people Nvidia is really hurting themselves with this slam of FutureMark. I used to really like Nvidia cards but lately they are really p!ssing me off. They need to stop making excuses and start getting a line of DX9 cards out there. WHERE IS THE GFX I STILL CAN'T buy one.

Looks like all of my system builds with DX9 cards are going to be ATI, not my fault.

Sazar

02-17-03, 12:25 AM

Originally posted by Myrmecophagavir
Bingo. I wish they wouldn't advertise it as "the gamer's benchmark

yah... hmm... perhaps...

but I REALLY REALLY wish nvidia would not put their ridiculous, Nvidia.. the way its meant to be played logo all over the place... :)

kyleb

02-17-03, 01:00 AM

i voted #1 like most so far, im not sure about the first part of the statement (not like any of us realy is one way or antoher), but most defenatly the second part is an issue. after all, not only are the cheaper parts lagging far behind the standards but we are still yet to see conformation of the nv30 running the bench well with anything but qestionable drivers.

as for "the gamer's benchmark" i agree that it can be missleading, but no more missleading than ati saying the "r300 core supports ssaa" or any vast number of examples i could think of. but in a situation like that no one is realy lieing by anymeans and the only way anyone gets mislead in situations like that is when they read too much into it, which is bound to happen. for instance, check out the_matrix's comments starting at near the botom of this is page for an example of an nvidia fan gone way too far:

case in point (http://www.beyond3d.com/forum/viewtopic.php?t=4277&postdays=0&postorder=asc&start=100)

i almost felt like signing up as someone brainwashed by ati and start an argument about how truefrom was going to rule the world.

digitalwanderer

02-17-03, 03:12 AM

Originally posted by Smokey
It also DOES not reflect game performance. The 8500 was faster in 3dmark than my gf3, yet in games it wasnt :rolleyes:

And my 8500 is faster in both 3dm2k3 & 3dm2k1se than my GF3 (but gets edged out in 3dm2k & futuremark99), and I also prefer my GF3 to game on. Why? Because it's better, smoother, faster, cleaner looking and it plays all the games I like a lot better.

Trouble is, my GF3 is starting to show it's age...especially with U2. :( It's weird, but the newer games (with the exception of GTA3, which is R*s fault) are starting to run better on my 8500 than my GF3. NOLF2, SoF2, & UT2k3 I'd really have to say might be nudging ahead of my GF3 in the actual gaming department...if you add in AA & AF it'd be a "gimme" to my 8500.

This test is supposed to be reflective of what games are going to be, and in that I don't think ANYONE can tell yet if it is or isn't. It's just a benchmark for now, I'm planning on not taking it too seriously yet.

Well, at least until some games come out that actually justify me upgrading my hardware to a level that will make this benchmark less of a slideshow on my system. Til then I'm planning on sticking with 3dm2k1se for a while longer as me DX bench of choice. It's a hell of a lot more reflective right now of the games I like to play. :)

Smokey

02-17-03, 08:46 AM

Originally posted by Myrmecophagavir
Bingo. I wish they wouldn't advertise it as "the gamer's benchmark", because maybe it's true that games don't use the same techniques (ie. PS 1.4). But if you view 3DMark as a test of what a card is capable of rather than trying to emulate gaming performance, it works better. 8500 can utilise PS 1.4 to get things done more efficiently than GF3, so why shouldn't it "win"?

It's like saying "The 8500 wins in app A, but the GF3 wins in app B, therefore app A is a bad benchmark for the card's performance". It's only that app A's programmers made an effort to take advantage of the 8500's extra features. If game developers would take advantage of it more widely then it would win in more tests!

Sorry I should have been clearer, I said 3dmark, meaning both 2001+2003 ;) My point was that fine in 3dmark, the 8500 is faster than my gf3, but im games its not.

Now digitalwanderer added some comments on this also, I dont have both cards myself, so its just from benchmarks on websites. But from what digitalwanderer said, the 8500 seems to be a bit faster now in newer games, this may have something to do with the faster core/memory speeds? Without looking it up, wasnt the 8500 core clocked at 275? my OC gf3 (non Ti) is clocked at 240, which is good for that core.

Back on topic, I dont think Nvidia would slam a DX9 benchmark, 3dmark or any other, just because they have new cards coming that dont support DX9 :rolleyes:

I havent been keeping up with the NV31 and NV34, but are they not going to support any new DX9 features :wtf:

silence

02-17-03, 08:56 AM

Originally posted by ZoinKs!
The new 3dmark is not made to measure performance in current games. It is made to predict performance in games which will be released 1-2 years from now.

A graphics card producing a high 3dmark03 score will be better in future games than a card with a low 3dmark03 score.

If you want to know performance in current games, then look at framerates in those games. If you're shopping for a new card and want one that will last, then look at 3dmark03 results.

1-2 years from now???........with only ONE PARTIAL dx9 test??do i have to remind you thet first test is dx7??

3dmark03 is more like for stuff we currently have plus little peak in future........

jbirney

02-17-03, 08:56 AM

Back on topic, I dont think Nvidia would slam a DX9 benchmark, 3dmark or any other, just because they have new cards coming that dont support DX9

Well if the rumors are true with the nV 31/34 being some subset of Dx9 features and with the weight that 3dmark2k3 gives the scores, we can almost bet they these nV cards will score pretty low.

Nutty

02-17-03, 09:24 AM

what are you on about?

the fx does what it is supposed to... its ps 2.0 does not seem as efficient as the 9700pro's for some reason @ the moment.. either because it is not or because of driver problems... dunno...

what problem exists is the dx8 gf4ti cards running the dx8 games with ps 1.3 instead of 1.4 hence using extra renderng pases and therefore garnering a lower score...

concerning comments about 3dmark03... READ THE WHITE PAPER BEFORE MAKING RETARDED COMMENTS

I was under the impression the PS2.0 shader results were gained from the DX9 scene. If I'm wrong tell me.

This "DX9" shader scene consists mainly of DX8 shaders. Presumably PS1.4. I just thought that maybe nvidia still dont expose 1.4 explicity, but thus promote 1.4 shaders to 2.0, and therefore incurr a performance penalty.

It was just a thought, chill out!

Uh huh. So I guess we don't really need DX9 at all just code everything in a 64k boundry and that's all it takes.

Eh? Theres nothing stopping you writing a DX9 demo in 64k. I'm just saying I wasn't very impressed with the demo itself. The ogre scene looked awful. That womans's hair was crap.

You know, why doesn't everyone read FutureMark's white paper on 3DMark03 it explains a lot and answers many questions and concerns people are having.

I have.

AND GET IT THROUGH YOUR HEADS that it is for FUTURE GAMES not what is being played now.

Thats the problem, Genius!! Future games WILL NOT use the crap algorithms used in this badly written demo/benchmark.

John Reynolds

02-17-03, 10:36 AM

Originally posted by Nutty
Thats the problem, Genius!! Future games WILL NOT use the crap algorithms used in this badly written demo/benchmark.

And exactly how do you know it's badly written? You've dissected the source code already? Or are you simply parroting a certain IHV's opinion?

Sazar

02-17-03, 11:48 AM

Originally posted by Nutty

Thats the problem, Genius!! Future games WILL NOT use the crap algorithms used in this badly written demo/benchmark.

there are many REAL games out there that unfortunately are also badly written :)

case in point... ut2k3... even though I play it flawlessly @ 1600x1200 epic really made me mad with the memory leaks and what not in the game :(

however it is still used as one of the premier benchmarking games out there... even with the splash screen... even with the memory leak issues... :) hence alls fair... init ?

btw I still don't understand why 3dmark03 is considered badly written...

tamattack

02-17-03, 12:34 PM

Originally posted by Nutty
I'm just saying I wasn't very impressed with the demo itself. The ogre scene looked awful. That womans's hair was crap.

Do you realize that each strand of the woman's hair is being rendered individually in that scene? Let me guess, you would rather have a big blurred texture there instead?

Sazar

02-17-03, 12:37 PM

Originally posted by tamattack
Do you realize that each strand of the woman's hair is being rendered individually in that scene? Let me guess, you would rather have a big blurred texture there instead?

yes it was quite impressive for a dx8 demo.. I was surprised @ the lighting/rendering/bump mapping...

noko

02-17-03, 01:49 PM

Part of DX9 ps2.0 is also ps1.4, 1.3 - 1.1. Meaning if parts of the benchmark is using ps1.4 is because that is all that is required for rendering that stage or even 1.1 version, it is all part of DX9 spec. That doesn't make it less then a DX9 benchmark. 3dMark03 supports most pixel shaders inherit in DX9 and uses the best one available for your card. Why penalize cards with ps1.4 support just because your card doesn't have it. Now some effects did require vs2.0 and ps2.0 to work reasonalbly where as ps1.4 and 1.1 with multiple passes wouldn't work. As for why FutureMark didn't use ps1.3? Well it just doesn't give you much more of anything is the bottom line.

Uttar

02-17-03, 02:41 PM

I voted option 1, but please note that it *isn't* exactly what I meant.

Option 1 suggests the NV31&NV34 couldn't run Game Test 4. I believe both can ( I'd probably really not be impressed by Game 4 performance, however... )

My opinion, however, is that NV34&NV31 performance has been cutted in the areas 3DMark 2003 find the most important.
I'd guess, for example, that if nVidia insists soo much that 3DMark 2003 does too much skinning, it's because NV34 performance for VS is simply crap.

Uttar

legion88

02-17-03, 03:56 PM

I pick option 5.

5) The benchmark did not provide the GeForce FX 5800U results that was anticipated either by themselves or by the public.

It does not mean that the benchmark was bad nor does it have anything to do with NV31/34 or current NVIDIA-based products.

This is nothing new to NVIDIA. NVIDIA has criticized Futuremark's products before (e.g. 3DMark2000 back in January of 2000). The results of 3DMark2000 was not inline with a certain reviewer's expectations (he's not exactly a tech-head). Rather than call the reviewer's bluff (i.e. he's an idiot), NVIDIA chose to attack the benchmark.

You will find, if you haven't noticed already, that reviewers have teflon shielding. They often don't get criticized for not understanding a benchmark's results or misusing a benchmark.