I found some interesting info about Aquamark3

[Hellbinder]

I was looking for another thread to post this in but i could not find one.

Anyway.. I had been trying to Figure out how the 5900U could be so competative with the 9800pro in this "Modern" Benchmark. Then I found this..

30 PS 1.1 Shaders

5 PS 1.4 Shaders

4 PS 2.0 Shaders

That is the actual count of the type and Number of Pixel shaders used in the Benchmark. Which of course ends up making sense why the Two cards peform so Closely together.

Which brings me to a little Side Question. Doesnt it seem a little odd that Nvidia is not Screeming about this benchmark for the same Reasons they hammered 3dmark03?

[Moo]

AM3 doesn't have nearly the name recognition as 3dMark.

[StealthHawk]

Hellbinder,

Is that the standard codepath, or a special NV3x codepath?

[StealthHawk]

Originally posted by Hellbinder
Which briungs me to a little Side Question. Doesnt it seem a little odd that Nvidia is not Screeming about this benchmark for the same Reasons they hammered 3dmark03?

No it doesn't. For the same reason why NVIDIA denounced 3dmark03 then continued to "optimize" for it. I think you know as well as I do, along with everyone else, than NVIDIA attacked Futuremark purely for PR reasons.

[digitalwanderer]

Originally posted by StealthHawk
Hellbinder,

Is that the standard codepath, or a special NV3x codepath?
Are we going to have control over which path a card runs regardless of what it wants or does the program choose based on the card and that's that...it's something I'm a tad confused on.

[bkswaney]

Do me a favor... Tell us what this means.
I'm sure there r many who do not know. :)

[MikeC]

The SVIST (Shader Visualisation Technique ) test indicates the type of pixel shader that's being used. When I ran SVIST on the 9800 Pro, I would say that 50-60% of the output was based on PS 2.0. I'm not sure if the test provides ending statistics since I cancelled it after 3 minutes or so.


http://www.nvnews.net/previews/AquaMark3/images/SVIST_more.gif


This information is from the AM3 documentation on shaders.


The effects and their variations are currently implemented by 190 different vertex shaders and 39 pixel shaders. These numbers of shaders in combinations with the various non-pixel shader rendering techniques for DirectX 7 class hardware allow for testing the rendering hardware and their resource management to its limits. In the following table you find an AquaMark3 shader overview.

Number of shaders:

VS1.1 (DX8) : 190
VS2.0 (DX9) : 3
PS1.1 (DX8) : 30
PS1.4 (DX8) : 5
PS2.0 (DX9) : 4

Average number of insturctions per shader:

VS1.1 (DX8) : up to 70
VS2.0 (DX9) : up to 50
PS1.1 (DX8) : up to 13
PS1.4 (DX8) : up to 13
PS2.0 (DX9) : up to 40


4.1.2 Floating Point Numbers In Shaders

AquaMark3 applies the whole range from FP64 to FP128 floating point accuracy. We think the developer is still in charge to decide which accuracy should be considered for a certain effect to optimize the rendering efficiency.

In general, every effect with floating point numbers covering a wide numerical range and the need to be represented by an exponent (as world space coordinates per pixel) benefits from a high internal precision (FP128). That means mainly special effects which use geometric information inside the pixel shader (as high dynamic range illuminated normal mapping with a self shadowing term) need high precision. AquaMark3 does not require functionality in this respect.

Nevertheless we use high internal precision for the texture coordinate iterators because AquaMark3 utilizes effects which calculate texture coordinates on the basis of world space positions (caustics, fog, detail maps etc.). Besides, the majority of AquaMark3s effects are implemented in less accuracy and make availible precision hints to the API.


There's a thread in the AM3 forum where this was being dicussed. It appears that AM3 will not "fall back" to an earlier shader version in the TRISCORE (default benchmark) run.

http://arc.aquamark3.com/forum/showthread.php?s=&threadid=101

[StealthHawk]

Originally posted by bkswaney
Do me a favor... Tell us what this means.
I'm sure there r many who do not know. :)

It means it's more of a DX8 benchmark than a DX9 benchmark.

One thing I'm still not clear on, is whether there are different codepaths for different IHVs. So far it sounds like there are. But I thought Massive said no IHVs received preferential treatment in their benchmark, which would make no sense if there were different codepaths of different cards.

[MikeC]

Originally posted by StealthHawk
It means it's more of a DX8 benchmark than a DX9 benchmark.


Hard to say at this point based on my previous comments on the SVIST test. AM3 may be more of a DX9 benchmark for graphics cards built around DX9 technology.

[bkswaney]

Is AM3 based off aquanox 2? Or is it totally new?

[The Baron]

Originally posted by bkswaney
Is AM3 based off aquanox 2? Or is it totally new?
AQ2 IS new.

[bkswaney]

Originally posted by The Baron
AQ2 IS new.


I know that B. :)

I was just wondering if AM3 was based off it or something even newer.
So AN2 uses DX9? I did not know that. I figured it was just a DX8.1 game. :)

[Sazar]

Originally posted by Hellbinder
I was looking for another thread to post this in but i could not find one.

Anyway.. I had been trying to Figure out how the 5900U could be so competative with the 9800pro in this "Modern" Benchmark. Then I found this..

That is the actual count of the type and Number of Pixel shaders used in the Benchmark. Which of course ends up making sense why the Two cards peform so Closely together.

Which briungs me to a little Side Question. Doesnt it seem a little odd that Nvidia is not Screeming about this benchmark for the same Reasons they hammered 3dmark03?

that information was posted in b3d.. and I posted it in another thread :)

however... I don't know if you read the follow up statement saying that even TR : AOD uses close to 14 ps 2.0 shaders in its benchmark... a far larger number than AM3...

I thought that was an interesting point though I cannot independently corroborate this :)

have to go by the word of the lads @ b3d...

mike has verified the number of dx9 shaders in AM3... and it does appear to be more dx8 than dx9 by all means which goes a long way to explaining why the performance delta between the r3xx and nv3x cards is not so big...

[Kihon]

Originally posted by Sazar
that information was posted in b3d.. and I posted it in another thread :)

however... I don't know if you read the follow up statement saying that even TR : AOD uses close to 14 ps 2.0 shaders in its benchmark... a far larger number than AM3...

I thought that was an interesting point though I cannot independently corroborate this :)

have to go by the word of the lads @ b3d...

mike has verified the number of dx9 shaders in AM3... and it does appear to be more dx8 than dx9 by all means which goes a long way to explaining why the performance delta between the r3xx and nv3x cards is not so big...

Dispite only having a few DX 9 shaders, if they are used more often than PS 1.x shaders, then it would still be a valid test of DX 9 performance. (I think).

The key thing is what the PS 2 shaders are doing.

[volt]

According to Massive, there are no IHV specific codepaths. 39 pixel shaders isn't a lot (that's including the ones from DX8)

[Dazz]

Originally posted by bkswaney
Is AM3 based off aquanox 2? Or is it totally new? Yes it uses the aquanox 2 engine to power the benchmark which was optimised for the nVIDIA cards.

[ReDeeMeR]

In other words, engine was built for Gforce 3, then update still keeping Nvidia in mind and now the benchmark is again 'optimized' for FX series heh

[Joe DeFuria]

Originally posted by MikeC
Hard to say at this point based on my previous comments on the SVIST test. AM3 may be more of a DX9 benchmark for graphics cards built around DX9 technology.

Unless, of course, certain DX9 cards choose to use DX8 paths....

[Toaster]

Why use ps2.0 when ps1.x is sufficient for the effect you want??

turning all ps1.x into ps2.0 would be rather pointless, the output would be the same. and you'd still need the ps1.x shaders for the dx8 cards.

besides, the amount of ps2.0 shaders doesn't mean anything, it's how often they get used whats important here.

[panzaman]

..why do you think Detonator 50 are out on the 15th in coincidence with Aquamark3??....guess what!!!;)

[Dazz]

nVIDIA have sure been busy in 'Optimizing' their drivers.

[StealthHawk]

Originally posted by Toaster
Why use ps2.0 when ps1.x is sufficient for the effect you want??

turning all ps1.x into ps2.0 would be rather pointless, the output would be the same. and you'd still need the ps1.x shaders for the dx8 cards.

The question is, could they have made better effects by using more PS2.0 instead of using PS1.x.

besides, the amount of ps2.0 shaders doesn't mean anything, it's how often they get used whats important here.

I have to concede that.

[Dazz]

From what i have seen PS2.0 adds more depth and realisim over PS1.x.

[aapo]

Originally posted by StealthHawk The question is, could they have made better effects by using more PS2.0 instead of using PS1.x.

In benchmark, which has an objective to test DX9 performance, this is not relevant. If it were a game, it of course would be crazy to use PS2 shaders too often.

The only problem with simple PS2.0 shaders in benchmarks is that they can easily be replaced by PS1.X shaders in the drivers, thus falsily improving performance. If I remember right, there once was a certain benchmark - wait it's name was 3dsomething03 or something like that - where the benchmark makers used too simple PS2 shaders that could have been replaced by a dishonest vendor.

But I bet it isn't easy to find a situation where you really need PS2 shaders. You all have seen the HL2 benchmark screenshots, and there is a minimal difference between DX9 and DX8 - which tells us that DX8 shaders are at least almost as good as DX9. If I've understood right, the only effect that can't be done properly with DX8 in HL2 is the high dynamic range light effects, because like the name says, there needs to be high range of intensity values in the shader code, which only DX9 FP numbers can offer.

I bet this is why things are generally more shiny and the water is 'brighter' in DX9 effects, there is more dynamic range available. But in the HL2 screenshot the DX8 actually looked better to me, the water was too shiny in DX9. In AquaMark3 the scenery with Radeons looked better and it had more dynamic range judging by the screenshots, IMHO, so mebbe the real DX9 is good for something after all. :D

[mrsabidji]

Originally posted by Toaster
Why use ps2.0 when ps1.x is sufficient for the effect you want??

Because the goal there is to benchmark the card's PS2.0 performance. Anyway, I agree with you concerning a real game situation...

mrsabidji