Sign in to follow this  
matches81

Questions about FXComposer's shader performance analysis

Recommended Posts

Hi there! Haven't been here a long time. I've recently started work on a small game project. Now it's at the point where I'll need a few shaders. So I downloaded FXComposer 2.5 and began coding. After my shader was done as a first draft, I started the performance analysis and have a few questions: a) I seem unable to get results for the GeForceFX series of GPUs. My shader (HLSL) uses ps_2_0 and vs_2_0, so I thought it should work. But all I get is a "CompileError" with no additional information. On GeForce6xxx and up all works fine. What do I have to consider to get this working? b) According to the analysis results, my shader uses 4 registers and 31 to 38 cycles on GeForce7xxx and below, but uses 38 registers and 133 cycles on any GeForce 8xxx. I'm confused as to why that may be. Any kind of info on that would be nice. If any additional information is required to help me, I'll gladly provide that. Just wanted to keep the original post short. I hope this is in the right section, was unsure where to put this. Thanks in advance for any help provided!

Share this post


Link to post
Share on other sites
The 7-series and 8-series have a radically different architecture, with the former having dedicated vertex and pixel processing units and the later having a large amount of scalar processing units that can operate on both. As such it's not surprising that the driver would create completely different microcode from the same ASM shader.

I have no idea why it won't work on the FX series, but I can't say I'm surprised either. That line of GPU's has some big problems, one of which is terrible SM20 shader performance.

Share this post


Link to post
Share on other sites
thx for the reply.
I knew the 8-series was a rather drastical change from previous GeForce GPUs, so my guess also was that these odd numbers were due to that. It's just that I don't know whether that's something I should worry about or if there's something I can do to change that.
Performance-wise the 8-series still is way ahead of the 7-series and below, as I would have expected. I guess it's more of a curiosity thing. If anybody has some more detailed info on this, it still would be appreciated.

GeForceFX performance analysis isn't that important, either, I guess. I was just curious why it wouldn't work at all. I know the FX had pretty abysmal SM2 performance, but it supports SM2, so it theoretically should work.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this