Performance issue with MBpro GeForce GT 750m + shaders vs Intel Iris Pro

I’ve been working on moving some particle systems and flocking systems to GPU using GLSL and i’m running into some speed issues that I don’t know how to resolve.

I’m running a new macbook pro with a GeForce GT 750m GPU. The problem is that when I run these programs using the integrated Intel Iris Pro, I’m getting framerates around 45fps, and when I switch to the NVIDIA card, I’m getting 15 - 20 fps max.

I’m not moving anything back from the GPU to CPU.

Has anyone suffered similar performance issues? Any thoughts on what to look at as possible fixes?

I can post code here if it would be helpful, although this is happening across many different examples of my own code, and even openCL examples taken from the NVIDIA website.

It might be time to break out the OpenGL Profiler and see what you can find. If you haven’t used it before, the OpenGL Profiler User Guide might be helpful and I’m sure some of the techniques were discussed in WWDC sessions too. Aside from that, vogl does run on OS X now and while Nsight Eclipse Edition runs on OS X, I think that is for CUDA debugging / profiling.

Blair Neal also wrote some stuff on the ogl profiler which might be helpful http://blairneal.com/blog/using-the-opengl-profiler-with-openframeworks-or-cinder-processing-etc-etc/

would be interested to hear what you find

Thanks for the advice, I’ll give it a shot and share any findings.