Fusion Not Using Nvidia GPU on Laptop

When using Fusion on my Win 8.1 Pro 64-bit laptop, it chooses the Intel graphics for OpenCL instead of my 2GB Nvidia 640m. This can not be corrected by forcing the use of Nvidia in the Nvidia control panel. This was a problem that myself and others were having with Fusion 7 also. Would be nice to see this corrected for Fusion 8.

Is it giving you a warning about that feature only being in Fusion Studio?

Of course, trying to manually select the OpenCL GPU in fusion gives that warning. It's understandable that selecting a specific GPU other than the one driving the program for the sole purpose of offloading OpenCL calculations should be a Studio exclusive feature. What I'm saying is that on Nvidia Optimus systems, the GPU driving the program should be the one that is used for OpenCL. Right now, if I tell Fusion to run with the Nvidia GPU from within the Nvidia Control Panel it will do so from what I can see in the Nvidia tool that shows what program it is currently powering. When I look in the Fusion Preferences though, I can see that Fusion is still using the Intel 4000 graphics for it's OpenCL.

You cannot select the OpenCL GPU from the Nvidia control panel. You can only select the OpenGL GPU. The application has nothing to do with it.

Selecting the OpenCL device is application specific, and in this case it is a Fusion Studio feature.

Yes, I'm aware of all that. What I'm getting at is developer intent. What was their intent when implementing this limitation? Once again, my impression was that this was meant to limit someone from offloading OpenCL calculations to a different card than was being used for GL. I'd have to go digging through past posts, but if I remember correctly, any BMD feedback on this issue indicated they were not aware of it and therefore I would think that having certain laptops always default to the iGPU no matter what was not the developer's intent and therefore a bug. If that was their intent, it would be good if they would state that somewhere so people are aware of it.

Personally, I think it's a silly limitation anyway. There are some OpenCL kernels that will be terrible on the Nvidia GPU and others that will be terrible on the CPU. This is increasing support burden for very little sales advantage.

Fusion lets you write and execute your own OpenCL kernels, but won't let you debug them on different OpenCL devices? That's just dumb. There, I said it.

I could understand them limiting the number of OpenCL devices, but even Fusion Studio only can use one.