New DX8.1 benchmark: Codecreatures

[Nexus]

From: http://www.codecult.com/ as seen on ECTS and GDC. You can download it here: http://www.codecult.com/benchmarks/Codecreatures_Benchmark_Pro_Setup.exe


1. A minimum of 512 MB RAM and Windows XP is
recommended (1 GB to measure pure graphics card
power) !
2. Requires DX8.1 (and ensure that Direct3D release,
not debug, is selected) !
3. Your graphics card should have hardware support
for Pixel Shaders !
(e.g. NVIDIA: GeForce 3 and GeForce 4 Ti,
ATi: Radeon 8500)
4. AGP Aperture size: This benchmark requires a total
video memory greater than 128 MB. This means that
if your graphics card has less than 128MB of memory,
then you MUST set your AGP Aperture size to 128MB
or greater. Anything less than this can result in
inconsistent results or stability issues.
5. Write protection must be disabled for all benchmark
files !

Care to post some numbers and/or screenshots? I have no DX8 card, not yet.

 

[Ascended Saiyan]

Beat me to it,but from what I gathered over at Rage 3d below 20 FPS seems to be the norm.

 

[LeStoffer]

Nice demo indeed, but I don't know how brilliant the benchmark is yet.

But let's get on with it, shall we?

Anyway, with a GF4 4600TI (yes I lost out to the dark side in the end!) Athlon XP 1800+, 512 MB and Win XP:

1024x768:
Avg frames/ sec: 29.9 (max = 38 )
Avg polys/sec 10.8 mio. (max = 16.2 mio)

1024x768 with 2XFSAA (enabled in Det drivers):
Avg frames/ sec: 25 (max = 31)
Avg polys/sec 9.5 mio. (max = 15.2 mio)

I couldn't run the official benchmark since my monitor doesn't run at 1600x1200, so...?

Anyway, it looks pretty indeed, but the tiny leaves on the trees does make a sort of "pixel popping" when in the distance (and 2xFSAA cannot make it go away totally). Secondly the demo is panning slowly around and overall looks smooth to me: Sometimes the camera "jumps" a tiny bit forward, however, but since the avg frame rate doesn't dip during this it could just be way the camera movement was recorded during the bechmark-setup. Don't know.

Anyway, I think that heavy use of advanced visuel features is okay if you use it for a slow moving game a la Myst. We don't need 60 FPS if the game restrict us from running around like crazy space cowboys, do we?

Hmmm. We need a lot of input before we can gauge whether this benchmark is going to be crucial or not. But it'll tell a truth or two, I'm sure.

 

[Dave Baumann]

This was on one of NVIDIA's list developers touting the GeForce4 at the launch. IIRC they said they took it as an extension of the original Chameleon demo. Anyway, I would suspect that GF3/4 would have been the primary development plaform for this demo - I'd be surprised if they used PS1.4 at all.

 

[2B-Maverick]

I only ran the Demo with sound on my 64MB Radeon8500 and got below 10fps average in 1280x1024. Dunno the exact number.. was between 7 and 9.
Will re-run tomorrow the whole range.
But since it needs 128MB of memory (kinda non-typical, eh?) i dont think my numbers will go over 10fps any time soon.

CU
Markus

 

[Doomtrooper]

This was on one of NVIDIA's list developers touting the GeForce4 at the launch. IIRC they said they took it as an extension of the original Chameleon demo. Anyway, I would suspect that GF3/4 would have been the primary development plaform for this demo - I'd be surprised if they used PS1.4 at all.

Unless this engine is falling back (and I don't think it is), and Geforce 3 cards can run it. This is another example again of not a Dx 8.1 benchmark but a Dx 8.0.
When reading the quote from Tomshardware:

THG: What's your view on NVIDIA's leaving out the (ATI) Pixel Shader of version 1.4?

Stefan Herget - Managing Director of Codecult: Thanks to its extremely modular and flexible architecture, the CODECREATURES engine is arbitrarily expandable and supports PS 1.4.
The game developer can implement special PS effects, therefore it depends on his decision whether version 1.4 or only 1.1 or 1.3 is supported. The library of prefabricated Shader effects provided by us will first be optimized for the common hardware base, and will thus use version 1.1 or 1.3, although the engine already supports 1.4. We will later extend our effect library directly for version 2.0.


They have chosen the Pixel Shader for best compatability IMO which is the ol' 1.1
So Again we are really not testing Dx 8.1 but 8.0 as Dx 8.1 requires PS 1.3-1.4 so Geforce 4's and 8500's should be the ONLY cards able to run this test.
I have fired off a email to confirm but I'll bet my Madonna Sex Magazine I'm right.

 

[pcchen]

A program written for DX8.1 won't work on DX8.0. It doesn't matter whether it uses pixel shader 1.4/1.3 or not.

 

[Doomtrooper]

Pixel Shaders are backwards compatible correct ?? So if the developer chose to support PS 1.1 then thats really not DX 8.1..correct ??
If this engine is using a higher pixel shader version then Geforce 3 cards should not be able to run them, and Ps 1.3-1.4 is Dx 8.1 specifications. If this engine is currently using PS 1.1, I can't see how it can be considered a DX 8.1 test when none of the advanced Pixel Shaders versions are being used.

 

[LeStoffer]

Pixel Shaders are backwards compatible correct ?? So if the developer chose to support PS 1.1 then thats really not DX 8.1..correct ??
If this engine is using a higher pixel shader version then Geforce 3 cards should not be able to run them, and Ps 1.3-1.4 is Dx 8.1 specifications. If this engine is currently using PS 1.1, I can't see how it can be considered a DX 8.1 test when none of the advanced Pixel Shaders versions are being used.

I understand your point Doomtrooper, but in DX 8.1 Pixel Shader Reference Microsoft states "pixel shader versions 1.0, 1.1, 1.2, 1.3, 1.4". You should look at DX 8.1 as DX 8.0 plus pixel shaders 1.2, 1.3, 1.4.

Anyway, if you haven't seen this already have a look at the DX 8.1 Pixel Shader Reference at Microsoft here:
http://msdn.microsoft.com/library/d..._vb/GRAPHICS/Reference/Shader/Pixel/Pixel.asp

In other words: I don't think we will ever see a benchmark/demo that only tests pixel shader version 1.3 or/and 1.4 (unless it's from nVidia or ATI which in both cases is, well, doh!)

 

[Doomtrooper]

Thanks for the link Lestoffer, truly confusing and frustrating as I see no point in advancing Pixel Shader revisions if nobody uses them.
The other issue I see here is the rules keep getting twisted...1st to be a Dx 8.1 compliant card you need PS 1.3 or 1.4...now we see because Dx 8.1 is backwards compatible game developers can use this backdoor and say the engine is Dx 8.1 yet really is only using Dx 8 hardware features. The major selling part of this engine is Pixel Shader support, they admitted the engine can do all versions so again why not have that option to do so if the work is already done.
I'll see what my email reveals later this week....

 

[pcchen]

Perhaps I should be more clear. D3D 8.1 is more than D3D 8.0 + PS1.2/1.3/1.4, although they are the most important new features (other new features are most about D3DX). Therefore, a program written and compiled with DX8.1 SDK won't work on a system with DX8.0 runtime. Of course, it is still possible to make a DX8.1 program to fallback to be compatible with DX8.0 runtime.

So to me a DX8.1 program means it requires (or at least utilizes) DX8.1 runtime. It may or may not work on a system with DX8.0 runtime. If a benchmark is intended for testing PS1.2/1.3/1.4, it should said "a PS1.2/1.3/1.4 benchmark" instead of "a DX8.1 benchmark." After all, DX8.1 includes PS1.2/1.3/1.4, and PS1.2/1.3 are very different from PS1.4.

 

[Matt]

I posted my results on the front page of 3DGPU, but if you haven't seen it yet, here's what I got:

P4 2.2GHz
512MB RDRAM
Visiontek GF4 Ti4600 @ 300/700

Total Official Score: 2377

1024x768: 29.8avg / 39max
1028x1024: 23.6avg / 30max
1600x1200: 19.1avg / 26max

Pretty sweet looking demo, like a glorified Nature demo. It'd be nice if, a year from now, games start looking like this.

 

[noko]

My results with a:

GF3 Ti200 @ 240/500 clock speed (Ti500 speed)
1.4gh T-Birdy
512mb of pc133 SDram

This is without antialiasing nor anistropic filtering

Seems like this game engine is way beyond the capability of my gig

I wonder since the polygon count is in the millions if the gf4s is being limited by AGP 4x transfer rates. Can anyone run the benchmark with a GF4 with AGP2x and then AGP4x?? I will later but I don't think in my case AGP is the limiting factor.

I can only dream now that future games will look this good when played. This is exciting times for me, playiny Astroids on a Atari console to this in 22 years is exciting to say the least.

 

[tEd]

Official Score: 2246
athlonXp1900
768MB 266DDR
asus gf4ti4600(ref. 28.32)

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

DETAILS:

average number of frames per second
1600x1200 -> 18.4
1280x1024 -> 22.5
1024x768 -> 27.4

max. number of frames per second
1600x1200 -> 25
1280x1024 -> 29
1024x768 -> 41

average number of polys per second
1600x1200 -> 7.7 mio.
1280x1024 -> 8.5 mio.
1024x768 -> 9.6 mio.

max. number of polys per second
1600x1200 -> 14.0 mio.
1280x1024 -> 14.0 mio.
1024x768 -> 13.8 mio.

total amount of frames
1600x1200 -> 2408
1280x1024 -> 2955
1024x768 -> 3604

 

[pascal]

I could not run it, my windows crashed for the first time restarting the computer without any warning (no blue screen)

My specs:
i815 b-step mobo (64mb AGP maximum aperture size)
256mb ram
P3-S 1.13GHz
GF3Ti200 64mb w/ Asus 27.42 drivers
Windows XP Pro

IMHO this benchmark specs requirement is too high
Where I can find screenshots? thanks
edited: I found two screenshots at 3DGPU. It looks impressive.

 

[Doomtrooper]

Perhaps I should be more clear. D3D 8.1 is more than D3D 8.0 + PS1.2/1.3/1.4, although they are the most important new features (other new features are most about D3DX). Therefore, a program written and compiled with DX8.1 SDK won't work on a system with DX8.0 runtime. Of course, it is still possible to make a DX8.1 program to fallback to be compatible with DX8.0 runtime.

So to me a DX8.1 program means it requires (or at least utilizes) DX8.1 runtime. It may or may not work on a system with DX8.0 runtime. If a benchmark is intended for testing PS1.2/1.3/1.4, it should said "a PS1.2/1.3/1.4 benchmark" instead of "a DX8.1 benchmark." After all, DX8.1 includes PS1.2/1.3/1.4, and PS1.2/1.3 are very different from PS1.4.

Thats my point Pcchen, the major selling point of THIS engine is the use of Pixel Shaders. The most hyped and significant part of Dx 8.1 is the NEW Pixel Shader revisions.
So its kind of misleading to people to say its truly a full compliant Dx 8.1 benchmark when the main feature of the engine is Pixel shaders and the vesion they chose was last years Dx 8.0 version.
Oh well ..

 

[pcchen]

If it requires DX8.1 runtime to run this benchmark, it is a DX8.1 benchmark. There is no so-called "fully compliant" DX8.1 benchmark, since there is no way (and nonsense) to use all functions provided by DX8.1. My point is, you can not equalize PS1.2/1.3/1.4 with DX8.1.

So I think it is fine to call a program "DX8.1 benchmark" even if it does not support PS1.2/1.3/1.4. There is no misleading here. One should not assume a DX8.1 program will always support PS1.2/1.3/1.4.

 

[Doomtrooper]

If it requires DX8.1 runtime to run this benchmark, it is a DX8.1 benchmark. There is no so-called "fully compliant" DX8.1 benchmark, since there is no way (and nonsense) to use all functions provided by DX8.1. My point is, you can not equalize PS1.2/1.3/1.4 with DX8.1.

So I think it is fine to call a program "DX8.1 benchmark" even if it does not support PS1.2/1.3/1.4. There is no misleading here. One should not assume a DX8.1 program will always support PS1.2/1.3/1.4.

Then there should be no such thing as compliant Dx 8.1 video card either

 

[Dave Baumann]

Pcchen,

While I agree that technically you are correct there I believe there is some validity in what Doom is trying to say. While the benchmark may require the DX8.1 runtime to operate I wouldn’t necessarily say that it’s a fully DX8.1 benchmark because it doesn’t make use of the features that differentiate DX8.1 from DX8.0. To my mind it’s a similar situation with Max Payne – while it requires the DX8 runtime to operate (because it does make use of a couple of minor DX8 features AFAIK) its hardly a showcase DX8 game because it still uses fixed function T&L and completely lacks any shading requirements.

 

[Reverend]

Maybe it's because some developers feel that there is more to DX8 or 8.1 than "mere" shading. DX8 provides for several strictly-performance related features compared to its predecessors. This could be more important for developers and I wouldn't blame them. Can't be close minded, y'know...