I have two Nvidia Cards installing in my computer running Ubuntu 12.04 LTS . Installing the drivers was so easy thanks to the active ubuntu community. I use these cards only for CUDA computation.
I would like to purchase a HD 7970 6gb because the extra memory allows me greater flexibility for my computations. I will be installing the proprietary driver from the AMD website.
Also, I plan to use only Python OPEN CL to program on the HD 7970 card.
My question is "Is it possible to have a mix of the Geforce and the Radeon HD cards in my system. Also, will there be any conflict with the two proprietary drivers ?
I have the latest nvidia drivers installed and for the Radeon, I will be installing the AMD catalyst 12.10 driver.
Any help would be appreciated. Please let me know if you'd like more information to better help answer this question. Also, since this is my first time posting, also let me know if this question is better suited for another forum .
I doesn't sure, but maybe there is a way to install only kernel module and CUDA implementation from nVidia driver (CUDA is enough, right?) without touching OpenGL and X.Org Server stuff (that will belong to fglrx).
Well, after a quick look through a catalyst package I only see minor obstacles.
For example /usr/lib/libOpenCL.so* will probably conflict with nvidias opencl libraries. But you could just put them in a different directory and use LD_LIBRARY_PATH or LD_PRELOAD when running programs on the radeon card.
If you only use it for opencl I think you wouldn't need the amd libGL.so either. If so, there is LIBGL_DRIVERS_PATH.
What could be problematic is having both the nvidia and fglrx kernel module loaded. I don't know if they conflict.
So it's probably a bad idea to use the default installation method of the drivers, but maybe someone has created a script to automate it yet.
Thanks a lot folks! That helped. I'll try to figure out how to resolve these openCL overwrites by different drivers. I have another question. Do I need to care if nvidia's libOpenCL libraries overwritten ? I will be using open CL only for the AMD card and using CUDA for nvidia.
What I am looking to do is have the NVIDIA card installed in the first PCIe slot and used as the Graphics renderer. I want the AMD card in the second slot and it used as the OpenCL compute engine.
Can the two proprietary card drivers reside in the kernel without conflict?
Does anyone know exactly how to end up with just the NVIDIA OpenGL stack to ensure graphics works on NVIDIA?
ChrisXY charaterized this as "minor obstacles", but a few more words of guidance would be appreciated.
ashteks did you ever get this to work?
I've been running a multiset setup with OSS Radeon drivers + Nvidia binary. The main desktop runs on integrated radeon with OSS and an clean instance of XBMC runs on the nvidia PCI card.
basicaly i did:
configure 2 setups on xorg.conf
reinstall xserver-xorg-video-radeon after installing nvidia's blob
point /etc/alternatives/x86_64-linux-gnu_gl_conf to mesa
export LD_LIBRARY_PATH and run xbmc on the second X with -layout -isolateDevice -modulepath (pointing to nvidia)