GPU from multiple vendors

Dawgmatix

Newcomer
I am trying to setup a box for opencl development. Would it be possible to have both an Nvidia and AMD gpu on the same machine? I only plan to use one of them for display, the other would only be used for compute. Primary goal for this setup is to have an iterative development environment where I can evaluate opencl kernel behavior on both implementations without having to spend on two development machines.
 
Yes, that will work. You will need to either make your software aware of how to select the right GPU, or disable OpenCL support for one of the GPUs at a time in the Registry, but either way that should work.
 
Why would it be necessary to disable OpenCL for one GPU? I can use OCL on my Intel iGPU or my AMD dGPU just fine in my personal box. Surely two dGPUs must work equally fine together.
 
Why would it be necessary to disable OpenCL for one GPU? I can use OCL on my Intel iGPU or my AMD dGPU just fine in my personal box. Surely two dGPUs must work equally fine together.
So that you don't accidentally start your OpenCL task on one GPU when you want the other one. If you know what you're doing with OCL queries this isn't a problem, but every now and then I find OCL software that just defaults to whatever the first adapter is, regardless of whether it's the one you want.
 
Righto, makes perfect sense from such a perspective... Strange tho that a programmer wouldn't take more care, considering that dual GPUs are quite common these days simply by function of the prevalence of iGPU CPUs.
 
Back
Top