Take the 2-minute tour ×
Ask Ubuntu is a question and answer site for Ubuntu users and developers. It's 100% free, no registration required.

I am exploring with the Toronto GPU Programming MeetUp whether I can add an ATI Radeon card for OPENCL GPGPU computing to my rig with an existing NVIDIA GTX 680.

The specific idea is to have a Radeon HD 7970 for OpenCL support, and a NVIDIA GTX 680 with binary drivers installed.

I understand that the binary driver works by replacing a lot of the X/OpenGL stack. I also know that there is work on adding OpenCL to the opensource Radeon driver & new gallium mesa stack, but that it is not done yet (and not compatible with the binary drivers anyway). I also know best practice when switching from one video card to another using binary drivers is to completely scrub the drivers.

But what I am not clear on as yet is the linkage between the Stream SDK and the X/OpenGL stack, such that I can use part of the Radeon driver for OpenCL support, but leave graphics to the NVIDIA stack.

Ideas that came to mind could include getting the the ATI driver to install to a different location, or installing the NVIDIA stack over the AMD stack (hopefully leaving the OpenCL specific parts alone).

Note that I have no intention of video/graphics on the Radeon card. It will only be used as a GPGPU.

I am looking for specific examples of documentation of anyone who has accomplished this. Google suggests is is possible with Windows 7, but my question is for Ubuntu 12.10 64 bit.

nouveau is not a solution for me at this time, as I am doing Linux based gaming on this rig and don't want to give that up.

If there is a better place to post this question, please let me know.

Thanks to anyone with ideas or suggestions on how to proceed.

share|improve this question
    
After some hardcore google-mancy it appears that this will work if the graphics card is AMD, and the Compute card is NVIDA. Install flgrx first, then install NVIDIA proprietary driver and tell it not to install the OpenGL libs. But what I am looking for is the reverse. If I can pull the packages apart, it may be possible to get libCL.so just from AMD and keep the NVIDIA OpenGL stack. But can the two kernel drivers co-exist in the kernel? –  Eric Carroll Feb 4 '13 at 4:14
    
Found the document on the NVIDIA install package components... –  Eric Carroll Feb 4 '13 at 4:22
add comment

1 Answer

Actually I managed to do just that, it was a year ago and I was mining some bitcoin with the Radeon card while having the NVidia card as the render card and for movie watching. I am not sure how I have done that but as I recall installing AMD drivers first and NVIDIA second will do just that. There was some instability due to the sequence in which the drivers were loaded. The AMD binary blobs worked only one way, either initialized before NVidia or after Nvidia - cannot recall for sure. Anyway, the point is, it is possible. Also, I might have installed NVidia drivers from repository (openSUSE) and AMD from installer.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.