3DMark03 - damage control for the GeForce FX?

[Kruno]

Originally posted by Hellbinder
Nvidias new Drivers *cheat* in 3Dmark03

http://discuss.futuremark.com/forum/showflat.pl?Cat=&Board=mo3dmark03&Number=1837817&page=4&view=collapsed&sb=5&o=0&fpart=1

Looks like the GFFX is winning becuase they are not rendering any explosions, and missing other effects. Guess thats how they got so fast so quick, yet made no change to the non scoring benchmarks...

Dont you find it a little odd that you somehow kyle didn't notice that? or make any mention of it? I find it pretty odd.

I love it when video cards don't render fx. :)

[Smokey]

Originally posted by Hellbinder
Nvidias new Drivers *cheat* in 3Dmark03

http://discuss.futuremark.com/forum/showflat.pl?Cat=&Board=mo3dmark03&Number=1837817&page=4&view=collapsed&sb=5&o=0&fpart=1

Looks like the GFFX is winning becuase they are not rendering any explosions, and missing other effects. Guess thats how they got so fast so quick, yet made no change to the non scoring benchmarks...

Dont you find it a little odd that you somehow kyle didn't notice that? or make any mention of it? I find it pretty odd.

Read the whole thread :rolleyes: I didnt see anyone posting that was using a GF-FX, did you? Also have you ever heard of BETA? Come back when you have something better to post, I mean your talking about beta drivers, and the most someone gained from those drivers was 200points, whoopie!

[Skuzzy]

From a developer point of view, I have considered all the current NVidia cards (sans NV30) to be DX7 cards.

They do not support PS1.4, which is a DX8 feature.

NV30 supports a subset of DX9. I don't understand why NVidia does this. It's not like they did not know about the specifications of DX9. Yet they add things that are not supported by DX.

Supporting a subset of the DX specs really creates problems for developerss using DX to move on to the next level of graphics programmability.

[StealthHawk]

Originally posted by Skuzzy
From a developer point of view, I have considered all the current NVidia cards (sans NV30) to be DX7 cards.

They do not support PS1.4, which is a DX8 feature.


PS1.4 is part of the DX8.1 spec. PS1.0-1.1 existed in the DX8.0 spec. this is one of the most ridiculous arguments i've ever heard, calling obvious DX8 hardware DX7. maybe if you want to call the gf4 a DX8 part instead of a DX8.1 part your argument could hold some water.

what about all the other DX8.1 cards? SiS Xabre doesn't support PS1.4, neither does Matrox Parhelia. i don't remember if Trident does either, but the point is kinda moot since they haven't released the card.

but seriously, why does nvidia take so much flak when no one other than ATI supports 1.4?

[Shinri Hikari]

Originally posted by Smokey
Read the whole thread :rolleyes: I didnt see anyone posting that was using a GF-FX, did you? Also have you ever heard of BETA? Come back when you have something better to post, I mean your talking about beta drivers, and the most someone gained from those drivers was 200points, whoopie!

You said what I was going to say.;)
Good post:cool:

[Shinri Hikari]

Since nVidia has pulled out of 3dmark, is this damage control thread moot?:D

[gokickrocks]

Originally posted by Smokey
the most someone gained from those drivers was 200points, whoopie!

you make it seem as though 200 points is easy to come by through tweaking (w/o oc'ing the core or mem) in 3dmark03

[Nemesis77]

Originally posted by StealthHawk
the point is that while they could be games, they aren't games. as has already been pointed out, even the Max Payne game test in 3dmark2001 didn't reflect real world Max Payne performance. so it's obvious that either the engines being used are not indicative of games, or the scenes being portrayed are not examples of real world games. either way something should be done to rectify this.

I still donät see a problem. All I see is NV whining about nothing. Here are the facts:

NV was big supporter of 3DMark01. They had exactly ZERO problems with it, even though it was not "real-life" benchmark.

Now, all of a sudden they have big problem with 3DMark03. Why? Because it uses PS1.4 that is in every way superior to PS 1.1, 1.2 and 1.3 (1.2 and 1.3 didn't bring any big improvements to PS1.1, PS1.4 did). And that really is NV's problem. They were the ones who chose not to support PS1.4. You could say that NV is holding the industry back by pushing inferior shaders to mainstream.

Ati is one generation ahead of NV in low-end and mainstream: their products all support DX9 and PS1.4, NV's products do not. That is the reason why NV whines. They made a wrong design-decision in the past, and that decisions has come around and bit them in the ass. Seriously, that is NV's problem, and no-one elses. Why should Futuremark cripple their software just so NV's crippled hardware would look better on it? The point of 3DMark is to show how the cards perform using latest and upcoming technology. PS1.1 that NV supports is old and inferior. PS1.4 and 2.0 are considerably better and there are titles on the way that take advantage of it.

I repeat: the point of 3DMark is to test performance in new and upcoming technologies. PS1.1 is neither of those. It's only used as an emergency backup if the card doesn't support PS1.4. PS1.2 or 1.3 are not used since the differences between those and 1.1 aren't that great. They had mostly trivial changes.

PS1.4 and 2.0 are the things that will be used in future games (Doom3 anyone?), and that's what 3DMark tests. If NV wants top keep on pushing yesterdays tech, they can do so. But they can't then whine if they don't look so good in softwre that takes advantage of new tech!

If NV wants to find someone to accuse over this thing, I suggest they look in to mirror. Fact is that Ati's entire product-lineup supports PS1.4 and DX9 (well, 9100 doesn't support DX9, but it has PS1.4). Large part of NV's lineup is still DX7 (GF4 MX)! NV has been holding the industry back, it's about time they catch up!

[Nemesis77]

Originally posted by abb
Well, I just ran the 3DMark 2003 on both, my Radeon 9700pro & my Ti4600. I scored an OK 5145 with my 9700 and a disgusting 1689 with my Ti4600.

And that's exactly how it should be. 3DMark03 is meant to test new tech (PS1.4 and 2.0 among others). GF4 does not support them, they have to use (slower) PS1.1 instead. And that REALLY hurts! But it's not problem with 3DMark, it's NV's problem for not supporting better tech.

Again: 3DMark is supposed to test new tech. GF4 doesn't support that new tech, so they don't do so well. It's NV's problem, not 3DMarks. Maybe NV should have decided to support new tech, instead of holding the industry back?

[abb]

Originally posted by Nemesis77

PS1.4 and 2.0 are the things that will be used in future games (Doom3 anyone?), and that's what 3DMark tests. If NV wants top keep on pushing yesterdays tech, they can do so. But they can't then whine if they don't look so good in softwre that takes advantage of new tech!

If NV wants to find someone to accuse over this thing, I suggest they look in to mirror. Fact is that Ati's entire product-lineup supports PS1.4 and DX9 (well, 9100 doesn't support DX9, but it has PS1.4). Large part of NV's lineup is still DX7 (GF4 MX)! NV has been holding the industry back, it's about time they catch up! Well said! heres a link and a quote from it that a Mod in another forum has pointed out to me:
http://www.extremetech.com/article2/0,3973,888205,00.asp

"We spoke with company officials from ATI and Dell, and both companies believe that 3DMark03's methodology and implementation are essentially sound. Dell, in particular, will be using the benchmark as one of its metrics for qualifying 3D GPUs.

So it seems that nVidia may just have to suck it up and live with whatever issues they have, since the biggest system OEM in the world will be judging its GPUs with this benchmark. " --Quote

For Nvidia to not support the benchmark because it makes their cards look like crap is a cheap shot and a desperate cry for help.
Abb

[Smokey]

Originally posted by gokickrocks
you make it seem as though 200 points is easy to come by through tweaking (w/o oc'ing the core or mem) in 3dmark03

Well yeah, I got an extra 100 points from just a driver update, ATI users got a big points increase with a driver update also. :rolleyes:

[gokickrocks]

the majority i have seen that updated to the new 3.1 drivers have lost a few points

[Kruno]

Originally posted by Smokey
Well yeah, I got an extra 100 points from just a driver update, ATI users got a big points increase with a driver update also. :rolleyes:

That's because they haven't really started increasing performance until now. :p

[StealthHawk]

does anyone have numbers for a r8500 or r9000 using PS1.4 vs PS1.1?

the way i see it, if the true DX8 gen cards from ATI gain nothing from PS1.4, there really is no reason to code in PS1.4. after all, from what i'm hearing, the r8500 (and by extension i assume the r9000 falls into this case) performs poorly, even with PS1.4 being used.

what this means is that the real cards benefitting in the DX8 games tests(ie GT2 and 3) are the DX9 generation of ATI cards, r9500 and r9700. and of course presumably the gfFX, as people say that if you support PS2.0 you also support PS1.4.

i really don't see why PS1.4 should be supported as a DX8 baseline if DX8 cards don't gain any benefit, so the conclusion really depends on the answer to my question. unless of course, PS1.4 is easier to code than PS1.1

[Kruno]

Originally posted by StealthHawk
does anyone have numbers for a r8500 or r9000 using PS1.4 vs PS1.1?

the way i see it, if the true DX8 gen cards from ATI gain nothing from PS1.4, there really is no reason to code in PS1.4. after all, from what i'm hearing, the r8500 (and by extension i assume the r9000 falls into this case) performs poorly, even with PS1.4 being used.

what this means is that the real cards benefitting in the DX8 games tests(ie GT2 and 3) are the DX9 generation of ATI cards, r9500 and r9700. and of course presumably the gfFX, as people say that if you support PS2.0 you also support PS1.4.

i really don't see why PS1.4 should be supported as a DX8 baseline if DX8 cards don't gain any benefit, so the conclusion really depends on the answer to my question. unless of course, PS1.4 is easier to code than PS1.1

Well if it is true that they render less polys in 3dmark then will that not benefit? :)

[Nemesis77]

Originally posted by StealthHawk
does anyone have numbers for a r8500 or r9000 using PS1.4 vs PS1.1?

Here (http://service.futuremark.com/servlet/Index?pageid=/orb/multicompare_view&p0=62135&n0=PS2.0+and+VS2.0&p1=59844&n1=PS1.4+and+VS1.1&p2=59140&n2=PS1.1+and+VS1.1&t=9700+Pixel%2FVertex+Shader+Comparison&k=10&s=101&l=1&c=366007170). Of course, that is R9700, but it gives you some idea regarding the difference in performance between different versions

EDIT: This is from Beyond3D forums:

Trolls lair dies on the 8500 with ps1.1, freezes after the loading screen and after rebooting I get the serious error message and one that the display adapter couldn't finish a drawing operation.

Athlon XP 12.5x141, 512mb DDR333, 8500 at 310/310.
With PS1.4 and VS1.1:

GT2: 7.3fps
Ragtroll: 5.0fps

PS1.1 and VS1.1:

GT2: 6.2fps
Ragtroll: 4.7fps

So there is a speed increase but its not a huge one.
PS1.4 performance seems to have gone up in the drivers recently though. In the 3dmark2001 APS test I used to get ~75fps on PS1.1 and ~85fps on PS1.4, I just got 102fps with PS1.4 with 1.1 still 75fps.

[Uttar]

Originally posted by Hellbinder
Nvidias new Drivers *cheat* in 3Dmark03

http://discuss.futuremark.com/forum/showflat.pl?Cat=&Board=mo3dmark03&Number=1837817&page=4&view=collapsed&sb=5&o=0&fpart=1

Looks like the GFFX is winning becuase they are not rendering any explosions, and missing other effects. Guess thats how they got so fast so quick, yet made no change to the non scoring benchmarks...

Dont you find it a little odd that you somehow kyle didn't notice that? or make any mention of it? I find it pretty odd.

Those drivers are 42.86
The drivers Kyle used were 42.67

So please, if you want to flame nVidia, at least get your facts straight.

I'm really getting annoyed by all of the crap being invented in this 3D Mark 2003 debate. I'm more and more considering writing an article about it. Anyone think I should, or is there enough articles about it already?


Uttar

[Kruno]

Uttar, you don't need to get emotional over companies. :)

[Smokey]

Originally posted by Uttar
Those drivers are 42.86
The drivers Kyle used were 42.67

So please, if you want to flame nVidia, at least get your facts straight.

I'm really getting annoyed by all of the crap being invented in this 3D Mark 2003 debate. I'm more and more considering writing an article about it. Anyone think I should, or is there enough articles about it already?


Uttar

I wouldnt bother Uttar, you can see from the forums that people believe what they want to believe :p