Nvidia and radeon working in same pc(i think Dithermaster said he was able to)

OpenCL
1

i heard that some people were able to do this. Dithermaster said he was successful on windows if so could he walk me through how he did or if anyone else has done it.

2

First time someone named me in a thread title <g>

I’ve had both NVIDIA and AMD cards in my HP Z820 for some time (different versions of each over time) as a way to confirm our kernels compile and work for both.

I didn’t do anything special except follow the HP guidance for which slots to use. I have a monitor plugged into each one; I’m not sure if this is required.

Each has current drivers; they don’t interfere with each other.

clGetPlatformIDs returns a platform for each, and within each platform the device represents the GPU. Of course you can’t create a context that contains both.

I often have the Intel CPU driver enabled as well, for a total of 3 platforms.

3

i new to all this and was looking for a way to use amd and nvidia gpus together. i noticed open gl had made hybrid pyshicsx but could your method work better instead. currently hybrid pyshicsx has been stalled and has not been updated for a substantial time and was interested if there was a better method

      • Updated - - -

i could find a way to ask you and your old thread was closed and was unable to pm u.

4

I haven’t used the physics packages so I’m not help there, sorry.

5

If you need GPU accelerated physics, then, I believe, Bullet Engine got OpenCL support. Check it out if using PhysX is not a requirement.

6

I’m very interested in building a machine with two OpenCL boards, but on Linux. The reason is that I need to do research on both CUDA and OpenCL. CUDA is available on NVIDIA only. And OpenCL is usually better supported on AMD than on NVIDIA.

Is it possible to do this on Linux? What approach would you take? Would it be better if one of the boards isn’t a gfx board and is a 100% GPGPU board without monitor output? Or is it better to have two gfx boards, both with monitors?

Also, any chances that current GPUs will be supported in Vulkan when/if it’s released?

Thanks a lot!!

7

[QUOTE=outofsync;39729]I’m very interested in building a machine with two OpenCL boards, but on Linux. The reason is that I need to do research on both CUDA and OpenCL. CUDA is available on NVIDIA only. And OpenCL is usually better supported on AMD than on NVIDIA.

Is it possible to do this on Linux? What approach would you take? Would it be better if one of the boards isn’t a gfx board and is a 100% GPGPU board without monitor output? Or is it better to have two gfx boards, both with monitors?

Also, any chances that current GPUs will be supported in Vulkan when/if it’s released?

Thanks a lot!![/QUOTE]

I use a workstation that has GPUs from NVIDIA and AMD as well as the pocl and beignet drivers installed. This works without any issues on Debian 8. I get 4 OpenCL platforms with each of the GPUs appearing in their respective platform and the CPU available in all but the NVIDIA platform.