ShaderMark NV40 vs R420

[Skinner]

When do you think we will see this new version? This month, next month, later?

Don't know. As it includes the latest HLSL compiler, I've to wait untill the dx9 sdk update (summer 2004) will be out of beta

Thomas

The beta is released, now I need a GF6800U :

The DirectX® Team is pleased to announce the release of DirectX 9.0 SDK Update
(Summer 2004) - Beta 3.

This Beta release contains the full DirectX SDK which features the pre-release
of the DirectX 9.0c developer runtime.
Areas of concentration in the DirectX 9.0 SDK Update (Summer 2004) Beta release
are:
- HLSL support for Pixel Shader & Vertex Shader 3.0
- Effects Framework performance improvements
- Pre computed Radiance Transfer improvements
- New Sample framework
- New & Updated Samples (and removal of some crusty ones)
- PIX tool for better debugging of Direct3D applications
- Introduction of the Preview Pipeline for easier content creation

 

[Pete]

the fillrate on both ati cards seems pretty low. is it a hardware issue? :/

Maybe it's a question of tweaking the programmable memory controller?

 

[Ante P]

I just tried the new beta and DX9.0c beta and now I can run PS3.0 in the tests with NV40.
All that happened though is that in a few test (4-5 of them) performance dropped. In the other tests it just remained the same.

 

[Dave Baumann]

:|

 

[Evildeus]

I just tried the new beta and DX9.0c beta and now I can run PS3.0 in the tests with NV40.
All that happened though is that in a few test (4-5 of them) performance dropped. In the other tests it just remained the same.

I supposed the performance is compared to PS2.0?

 

[Bjorn]

I just tried the new beta and DX9.0c beta and now I can run PS3.0 in the tests with NV40.
All that happened though is that in a few test (4-5 of them) performance dropped. In the other tests it just remained the same.

Performance dropped by how much ? (still a beta though, perhaps for a reason)

 

[Ante P]

I just tried the new beta and DX9.0c beta and now I can run PS3.0 in the tests with NV40.
All that happened though is that in a few test (4-5 of them) performance dropped. In the other tests it just remained the same.

Performance dropped by how much ? (still a beta though, perhaps for a reason)

In some tests it dropped down to 50% of what I get in PS2.0b.

In any case I just got word from nVidia that they have a new driver on its way with loads of fixes (high quality/ trilinear opt off working again etc.) so I'll just wait for that one before I publish anything.

 

[Bjorn]

In some tests it dropped down to 50% of what I get in PS2.0b.

Ouch.

In any case I just got word from nVidia that they have a new driver on its way with loads of fixes (high quality/ trilinear opt off working again etc.) so I'll just wait for that one before I publish anything.

High Quality = no angle dependancy ?

 

[Ante P]

High Quality = no angle dependancy ?

Nope, just the same filtering quality on all texture stages.

 

[Bjorn]

High Quality = no angle dependancy ?

Nope, just the same filtering quality on all texture stages.

That was what i suspected

 

[Evildeus]

In some tests it dropped down to 50% of what I get in PS2.0b.

Why with PS2.0b? :? You are comparing a X800 with a 6800?

 

[DemoCoder]

Did you check to see what FXC is generating? It still seems to have some issues.

 

[Ante P]

In some tests it dropped down to 50% of what I get in PS2.0b.

Why with PS2.0b? :? You are comparing a X800 with a 6800?

yeah well ps2.0 and ps2.0b is a very slight difference in any case (performancewise in this specific test that is)

 

[Evildeus]

Can you tell me more in detail what you are comparing with? 6800 on PS3.0 vs X800 on PS2.0b? that's right? Why not with 6800 on PS2.0?

 

[Ante P]

Can you tell me more in detail what you are comparing with? 6800 on PS3.0 vs X800 on PS2.0b? that's right? Why not with 6800 on PS2.0?

which comparison?
in the review I do 2.0 vs 2.0 currently
I will do 2.0a vs 3.0 or 2.0b in the follow up

the 50 % drop was referring to 6800 foing from 2.0b to 3.0

 

[Evildeus]

Ok, that wasn't clear, thanks.

 

[LeGreg]

the 50 % drop was referring to 6800 foing from 2.0b to 3.0

my guess the test doesn't do exactly the same depending on the profile it is compiled to.

Have you compared the output ps3.0/ps2.x asm ?

Does it use _pp in the ps3.0 case ?

 

[Ante P]

the 50 % drop was referring to 6800 foing from 2.0b to 3.0

Have you compared the output ps3.0/ps2.x asm ?

Does it use _pp in the ps3.0 case ?

I haven't used any tool (like compressionator) but they seem to produce adentical output

as for _pp you can choose wether you want it or not like for all the other modes

 

[Dave Baumann]

He's asking if you've taken a dump of the compiled shader assembly code to compare if the output of the SM3.0 compiled version is the same as the SM2.0 version.

However, is it not the case that ShaderMark just uses the HLSL profile selected, in which case it will be HLSL producing different output from for the SM2.0 and SM3.0 versions.

 

[Ante P]

He's asking if you've taken a dump of the compiled shader assembly code to compare if the output of the SM3.0 compiled version is the same as the SM2.0 version.

However, is it not the case that ShaderMark just uses the HLSL profile selected, in which case it will be HLSL producing different output from for the SM2.0 and SM3.0 versions.

Ahh sorry read too quickly.

Yes you're right.