Archives Discussions

[jammer]

Hi, I'm using GPUPerfStudio to debug my OpenGL applications and everything ran flawlessly until I upgraded Catalyst drivers from 10.8 to 10.11. Suddenly, GPUPerf is running very slowly, taking almost minute to connect to application or react to user actions. When I select HUD controls it even crashes.

When I rolled driver back to 10.8 all issues dissapeared, but I need new drivers due to OpenGL 4 support.

My system: Win7 32-bit, Radeon 5970 2GB, Core i5, 4GB RAM

Tried this also on Win7 64-bit with Radeon 4650 1GB, Core2duo,4GB RAM - exactly the same problems

EDIT: I've tried GPUPerfStudio 2.2 and 2.3, same issues

Thanks for help

[plohrmann]

Hello,

Sorry for the delay in responding. We are aware of the issue and a bug has been filed against it. The only work around in the short term is to use the Settings dialog to increase the message timeout. Although I haven't done any investigation into it yet, I believe it is related to the maximum number of textures that are available in OpenGL 4 and a non-optimal communication between the server and the client about those textures.

As such, I expect that you'll find the "HUD Controls" and "All Textures" features to experience this slowness (in addition to the connection), but other features, such as "Active Textures", should be quick.

To help guide our OpenGL development and support, can you please let us know what features of OpenGL 4 you are currently using?

Thank you!

[jammer]

Hi,

Thanks for the answer. I don't think it's OpenGL 4 fault only, because I have the same problems under OpenGL 3.3 with Radeon 4650.

As you say, slowest operations are connecting to application and enabling HUD controls, other functions (active textures or shader code) run rellativelly fast.

Currently, I use most of the OpenGL 3.3 functions with a few OpenGL 4 things (textureGather) but I'm planning to use more features (such as tesselation) in the future.

[jammer]

Still no improvement?