Late, noisy, HUGE!!! - makes 17k Mark...

[Neeyik]

No, I am blaming Futuremark for being *inconsistent* in this regard. SE proves this point with it's Advanced Shader test. Whether or not to add this test to the final overall "score" is a totally different matter. But the fact remains the Nature test *does* have a non-DX8 mode of operation, but this mode was not chosen to be allowed, tested or scored.

Out of all your lines of thought, this is the one that still puzzles me the most. What non-DX8 mode are referring to with the Nature test? Are you talking about the Nature part in the demo? The two are different - there is no lake material in the demo, since that there aren't many cards that are able to do per-pixel cubic bump mapping without pixel shaders. The whole point behind the final game test is to use pixel shaders - if you go through each game test, there is an increasing level of "DX8-iness" about them. The use of vertex shaders and vertex blending changes as you go through each of the tests, with the final one (Nature) being the most DX8-y as it uses plenty of vertex shaders, plus the heinous pixel shaders.

SE takes a 180 degree change in ethic and adds a DX8.1 test, doesnt code this even to the level of DX8.1 standards, provides a fall-back mode for non DX8.1 compliant hardware AND scores them.

A DX8.1 feature test was added, not a scoring test. All the entries in the database are unaffected by this (which was the whole point of not changing the game tests in SE). There's a fallback in the APS feature test, simply because it's possible. As everyone is aware, you can "do" PS1.4 stuff with 1.1 routines but there is no fallback from PS1.0.

Spend about 6 months in the SDK and repeat the above statement and see if you can do so with a straight face. It's also funny you are assuming that anything coded in PS1.4/single pass will obviously perform *lesser* or equivalent to multi-pass/PS1.1 (as the Advanced Shader test propagates). There are enough developer quotes around to prove this is NOT the case in reality, but not as illustrated by 3DMark2001SE.

Sorry - entered late into this game. Too busy learning C++ at the moment; when I'm happy enough then I'll start with the SDK. Where have I assumed anything about PS1.4-1.1-blah-blah though? Where did I say that I believe that a scene rendering 1.4 routines will be equal to or slower than multipass 1.1? I said that you're assuming that there should be a huge difference in the way that a Radeon 8500 handles the APS test compared to a GF3 (which there isn't), by the way you were claiming that it is carefully coded in such a way to not have any real benefit for PS1.4 cards, and in fact has a "fallback" mode, which is tested and performance added for non DX8.1 compliant hardware."

Edit: typos

 

[K.I.L.E.R]

it is carefully coded in such a way to not have any real benefit for PS1.4 cards, and in fact has a "fallback" mode, which is tested and performance added for non DX8.1 compliant hardware.

Stop pounding out conspiracies. I get more than double the framerate using PS 1.4 than PS 1.1 on my Radeon.
(refers to Ati techdemos, 3dmark 2k1 and some other various non-at-madonion demos)

 

[Doomtrooper]

Worm,

I don't consider you guys as the 'bad guys', I've always enjoyed reading your posts yet there is legit reasons why I've never loaded the SE version (I still have the original) and why people thought the SE version was a joke.

You advertise your Advanced Pixel Shader as advanced yet still allow PS 1.1 users to see it (fall back option)..whats so advanced about it ?? This makes no sense at all, there is tests made that only Ps 1.4 and up could do with any speed

I read Neeyiks post about comparing 8500 scores in Nature and for the life of me I can't figure out what is so difficult to understand, there is no reason Nature could not have been written with PS 1.4 support, do you think 8500 card users would have enjoyed looking at the difference single pass vs multipass...and would be very easy to compare scores.

Your own PR states DX 8.1 benchmarking yet the only DX 8.1 test is not even allowed to score..makes no sense at all.

So to put is simply Madonion/Futuremark let down all the 8500-9000 card owners by simply not exposing the superior Pixel Shader version in a better light, we were left with a screensaver with fish going through it (btw that any PS 1.0 card could still see) which I could get by turning on my screensaver.

Again Nature not supporting DX 8.1 PS 1.4 was a extreme let down, and how can anyone say there is not something 'fishy' about that.
I am still waiting for a legit answer to why Future doesn't support PS 1.4, the only thing that would change would be the 8500 cards would gain some frames, why is that so bad

 

[K.I.L.E.R]

Again Nature not supporting DX 8.1 PS 1.4 was a extreme let down, and how can anyone say there is not something 'fishy' about that.
I am still waiting for a legit answer to why Future doesn't support PS 1.4, the only thing that would change would be the 8500 cards would gain some frames, why is that so bad

Please, I beg of you to let it go.
There is no conspiracy here. 3dmark 2k1 is old and I don't see why they should bother adding features to it especially since they are concentrating on 3dmark2k3.

 

[Sharkfood]

Are you talking about the Nature part in the demo? The two are different

followed by "Advanced Feature Test"-

A DX8.1 feature test was added, not a scoring test.

You are on the outer cusp of understanding the difference.

How can it be a DX8.1 "feature" test if it uses no DX8.1 "features"? (i.e. single-pass, PS1.4 shaders). It does, but on a non DX8.1 card, it doesn't.. Hmm.. So how is this any different?

Put aside the Nature test and it's inclusion in the overall, final score for just a moment. There are reasonable debates for this (ORB result sorting, etc.etc.) but all come down to trying to predict the intention of the test, which can't be scientifically proven.

Now back to the Advanced Pixel shader test - it does provide a "score" - but this isnt tabulated into the final, overall 3dmark score. Getting back to the "Hmm.. It's a test for DX8.1, but yet it is coded in such a way to be easily reproduced on the lowest common denominator- ala DX8.0/PS1.1." And moreover, it is produced and scored (but not final score inclusive) on non-DX8.0 hardware.

The problem is- there is nothing in DX8.0/8.1 that cant be brute-forced in DX7.0. It's a simple fact. What can be done in shaders, single pass or hardware can also be done in software. This usually defines the performance (i.e. software/emulated might be 1fps versus 70 fps in DX8.1). Obviously the level of achievement when *compromised* for performance can also dictate how something is designed. The "advanced shader" test is a shining example of this.

Had The Onion chosen a true, PS1.4 style test- there would have been some margin of improvement illustrated by PS1.4. Something similar to the "Ocean" screensaver now popular amongst ATI owners. Something with both a water surface, cloud cover and sky. Something that actually exercises what PS1.4 can do. Unfortunately, with such a test- it wouldn't have the same false message the test current yields- which is a PS1.0/1.1 hardware can perform equal to or exceed a PS1.4 hardware.

I have no issue with this different take on the benchmark. I will have a problem if 3dmark03 will suddenly decide to take the "Nature" ethic if it better suits the IHV the last four or five changes in ethic over the years have illustrated. Consistency is the important element.

So in conclusion, it's simply an overwhelmingly well supported foundation to insist SOME level of consistency be present in the next version of 3DMark. For how MadOnion has approached DX benchmarks, the ethic, ideology, and design approach was defined with 3dmark2000, changed for 3dmark2001, then changed yet again for 3dmark2001SE. Whether you want to account these changes to sheer coincidence or factor what singular IHV benefitted from each occurance is totally up to you.

I simply ask if the ethic/approach/ideology for the result in 3dmark03 be taken into account with the past examples, and if any amount of objectivity is to be reached, it surely should AT LEAST follow the trends/standards set forth in 3DMark2001SE involving both the Nature test and the Advanced Shader tests.

 

[Neeyik]

To Doomy:
APS is "advanced" because it is doing more "PS stuff" than the original PS test.

I should imagine that FM/MO wanted the scores from 3DMark2001SE to be directly comparable to those from 3DMark2001 - otherwise its another long wait to build up data for measurement services. To keep the scores comparable means not changing the scoring tests.

Actually, as Sharkfood has pointed my lack of understanding, it would be nice to know from somebody whether a 4 texture stage cubic bump mapping routine using PS1.0 would be noticably slower than if you used 1.4 code.

 

[Nick[FM]]

But the fact remains the Nature test *does* have a non-DX8 mode of operation, but this mode was not chosen to be allowed, tested or scored.

You make me want to cry..

There's no use talking to you. You have no idea what you are talking about, and all your "proof" is only something you have made up in your head.

*MEDIC*

 

[Neeyik]

Getting back to the "Hmm.. It's a test for DX8.1, but yet it is coded in such a way to be easily reproduced on the lowest common denominator- ala DX8.0/PS1.1." And moreover, it is produced and scored (but not final score inclusive) on non-DX8.0 hardware.

Whoa horsey! "Scored on non-DX8 hardware"? Which ones? What non-DX8 hardware can run the APS test? me = puzzled

You are on the outer cusp of understanding the difference.

Naw, don't believe in that astrology stuff...

 

[Tahir2]

LOL.. On a different note I just spotted your signature worm, so whena re you going to add a link to your new 3DMark03 score?

 

[Althornin]

]

But the fact remains the Nature test *does* have a non-DX8 mode of operation, but this mode was not chosen to be allowed, tested or scored.

You make me want to cry..

There's no use talking to you. You have no idea what you are talking about, and all your "proof" is only something you have made up in your head.

*MEDIC*

Actually, he does have a point.
i only wish you would try to see it, instead of postng melodramatics.

Look - an "advanced Pixel Shader" test with a fallback to non advanced modes is kinda of ...worthless?
It also represents a switch in benchmark ethics from the nature test.
There have been several switch like this in 3dmark history, and ALL of them seem to favor nVidia.

Now, i am not saying there is a conspiracy, but it does seem to point a finger, thats all.
I am also aware that these changes could be just coincidence, but that seems unlikely.
I dont think you guys consciously think "lets boost nVidia and screw other IHVs" but i do think it has become an unconscious thing that just happens.

Or, you can explainb why shark is wrong, instead of say "OY vay! get me a medic".

 

[Doomtrooper]

Neeyik how could it not be measured ?? i.e 8500 gets 50 fps in 3Dmark 2001 and 70 fps in 3Dmark 2001 SE..or better yet SE version gives a fallback option for PS 1.1 (pop up..whatever)..its done on the Advanced Pixel Shader test so 'it couldn't be done' is not a excuse.

 

[nggalai]

Hi Sharkfood,

Neeyik-
I dont subscribe to the "links plz" teenage moronic diatribe fanclub, thanks.

I prefer "facts" as in "can be reproduced by anyone that takes the time to actually put the proof to the test" rather than "website fanboi #1 with FrontPage can specialized create."

No offence, but you said "No rumors.. All documented fact." Documented implies, well, that claims such as, quote, "A test included in the score that can only be run on one IHV's product AND added to the final score to ensure other IHVs are artificially penalized by score" are documented--in this case, that you can provide a document that clearly states that the test was included to "ensure other IHVs are artificially penalized by score." I don't see how such a "fact" can be reproduced--this is clearly a case of "show us some proof," or, rather, an implication. Not a "documented fact."

ta,
-Sascha.rb

 

[Nick[FM]]

Worm,

I don't consider you guys as the 'bad guys', I've always enjoyed reading your posts yet there is legit reasons why I've never loaded the SE version (I still have the original) and why people thought the SE version was a joke.

People in this case my dear friend is you, couple of your buddies and Sharkfood.

You advertise your Advanced Pixel Shader as advanced yet still allow PS 1.1 users to see it (fall back option)..whats so advanced about it ?? This makes no sense at all, there is tests made that only Ps 1.4 and up could do with any speed

Err.. Not sure how to explain this so that you get it.. Ok, we call that test "advanced pixel shader test" because it uses PS1.4, if the hardware is capable to render it. If not, then it uses PS1.1. I'm not sure how to get thru to you.. You don't seem to understand it even if I would explain it in Finnish.. It is advanced since it uses PS1.4, compared to simply using 1.1.

I read Neeyiks post about comparing 8500 scores in Nature and for the life of me I can't figure out what is so difficult to understand, there is no reason Nature could not have been written with PS 1.4 support, do you think 8500 card users would have enjoyed looking at the difference single pass vs multipass...and would be very easy to compare scores.

Your own PR states DX 8.1 benchmarking yet the only DX 8.1 test is not even allowed to score..makes no sense at all.

*sigh* I said that we didn't want to start editing the tests. If we would have put in PS1.4 support in the Nature test, it would most probably have changed the scoring + it might have changed the code in ways that we didn't want to -> affecting the PS1.1 compliant cards -> resulting in not comparable results between 2001 and 2001 SE. I don't know all the technical reasons, but I know that there were many concerns. So we did an extra test to test the performance of PS1.4.

It is DX8.1 benchmarking. SE has the advanced PS test, which uses PS1.4 (only found from DX8.1+).

So to put is simply Madonion/Futuremark let down all the 8500-9000 card owners by simply not exposing the superior Pixel Shader version in a better light, we were left with a screensaver with fish going through it (btw that any PS 1.0 card could still see) which I could get by turning on my screensaver.

As far as I know, the only superiority the PS1.4 has over 1.1 is that it can do some stuff in 1 pass, whereas the 1.1 needs to go in 2 or more passes. Correct me if I'm wrong. So, the Advanced PS does just that. You can compare 1.4 vs. 1.1 if you like. What's wrong with that?

Again Nature not supporting DX 8.1 PS 1.4 was a extreme let down, and how can anyone say there is not something 'fishy' about that.

I explained this a bit up. Please check that. There's actually nothing fishy at all. Use your head.

I am still waiting for a legit answer to why Future doesn't support PS 1.4, the only thing that would change would be the 8500 cards would gain some frames, why is that so bad

For the 1000th time, we do support PS 1.4 in 3DMark2001 SE.

 

[Neeyik]

Doomy - I'm not suggesting that it couldn't be done, but that it perhaps it wasn't a good idea at that time. The ORB database would have 9 months of 3DM2k1 data for the 8500 (R200 released in October 2001, SE released in June 2002); lots of useful info for the other FM services. Change the tests so that the scoring ends up different or whatever and then you've got another lengthy process of building up data again.

 

[LeStoffer]

Look - an "advanced Pixel Shader" test with a fallback to non advanced modes is kinda of ...worthless?

You have to remember that DX 8.1 does not equal PS 1.4 support. DX 8.1 was about expanded pixel shader functionality with the versions 1.2, 1.3, and 1.4.

While we might find PS 1.4 more advanced, you cannot really IMO follow DX 8.1 PS-specs by making an engine which only have support for the highest level of pixel shader. Go ahead and make an IHV demo, but for an independent software firm? I don't think so...

BTW: If created fairly the APS is not worthless at all, as it should display performance differences between the different PS.

 

[Nick[FM]]

Actually, he does have a point.
i only wish you would try to see it, instead of postng melodramatics.

Trust me, I have posted about this too many times. That's why..

Look - an "advanced Pixel Shader" test with a fallback to non advanced modes is kinda of ...worthless?

Well, not the way I see it. I think the Advanced test is actually very good. You can compare the PS 1.4 vs PS 1.1, or even PS 1.4 vs. 1.4. It is all "advanced" compared to the normal PS test. See my point?

It also represents a switch in benchmark ethics from the nature test.
There have been several switch like this in 3dmark history, and ALL of them seem to favor nVidia.

Oh dear.. How many times do we have to go over this? The feature tests do not affect the score. We can add 1000 feature tests in a patch if we want. But we must not tamper with the game tests. We can not risk that the whole score system goes nuts. We wanted to keep 2001 and SE comparable, and so we didn't touch the game tests. If we would have, there would have been a risk that the new code would have changed the whole score system, and .. you know what I mean. If we would have added 1 new GT using PS1.4, it would also have changed the scoring. There were no good solutions, so we made the introduction of PS1.4 into a feature test.

Now, i am not saying there is a conspiracy, but it does seem to point a finger, thats all.

Didn't mama tell you that it is rude to point fingers at others?

I dont think you guys consciously think "lets boost nVidia and screw other IHVs" but i do think it has become an unconscious thing that just happens.

The reason for this discussion goes way back to 99. It started when we introduced HW T&L. It's an old conspiracy theory from the old days that some .. eh.. "fellows" have been feeding year after year. Nothing more, nothing less.

Now I need to do some real work again. I'm out of this discussion until something new and interesting comes up. Anything special, then email me.

/lurking mode: ON

 

[Nick[FM]]

BTW: If created fairly the APS is not worthless at all, as it should display performance differences between the different PS.

But it does. You can compare 1.1 vs. 1.4 with it.

 

[Althornin]

]

BTW: If created fairly the APS is not worthless at all, as it should display performance differences between the different PS.

But it does. You can compare 1.1 vs. 1.4 with it.

no it doesnt.
And thats the problem. The Advanced shader test does NOTHING that gains from ps 1.4
and is worthless because of that.
Advanced PS !=1.1
I never said it needed to be 1.4 only, but having a fallback to DX8.0 is stupid.
I'd quote your response to mine, but you didnt actually answer any of my questions, and only misinterpreted what i said.
You keep coming back to the argument "the feature test isnt int he final score" - DUH?
We know this! The problem is, the feature test DOES recieve a score. A score that is worthless for the reasons outlined in this post. A score that favors nVidia hardware.
So in the end, nevermind, i cant seem to get you to even TRY to look at this issue from an outside perspective and explain why it appears the way it does. Enjoy your lurking.

 

[Doomtrooper]

]
It is DX8.1 benchmarking. SE has the advanced PS test, which uses PS1.4 (only found from DX8.1+).

Yes it uses Pixel Shader 1.4, BUT also looks the same on Pixel Shader 1.1 so again I ask whats advanced about it that is allows the older version to look the same.
According to ATI on their Ocean screen saver PS 1.1 couldn't render it with any speed, there is two Pixel Shaders being used, the water and the sky..something similar should have been done here.

I explained this a bit up. Please check that. There's actually nothing fishy at all. Use your head.

I am using my head ,I don't consider a DX 8.1 requirement being put with EMBM 'supporting it'...sorry.

For the 1000th time, we do support PS 1.4 in 3DMark2001 SE.

If DX 8.1 PS 1.4 is a feature then so is DX8.0 PS 1.1 so remove nature from scoring.

 

[Nick[FM]]

I'm so bad at only lurking.

]

BTW: If created fairly the APS is not worthless at all, as it should display performance differences between the different PS.

But it does. You can compare 1.1 vs. 1.4 with it.

no it doesnt.
And thats the problem. The Advanced shader test does NOTHING that gains from ps 1.4

But it does. Here's a quote from the help file:

The same effect can also be achieved using Pixel Shader 1.0, but then rendering the water surface requires two passes. Graphics hardware that supports Pixel Shader 1.4 (or higher) render the water in a single pass.

So you mean that it doesn't speed up the scene by doing it in 1 pass?

The problem is, the feature test DOES recieve a score. A score that is worthless for the reasons outlined in this post. A score that favors nVidia hardware.

So, basically what you mean is that eventhough some cards can only do PS1.1 and does the water surface in 2 passes, it still favours them? If I was confused, now I'm lost!