I’m working with 3D graphics and Autodesk Maya on Linux. In my workstation has two GPU NVIDIA Quadro, model K2000 and K4000. Two displays are connected to the Quadro K2000, and Quadro K4000 is used as an additional buffer memory and co-accelerator at NVIDIA CUDA or OpenCL.
On a Windows the default OpenGL driver allocates processing on both GPU. That is memory and computation of the virtual space are performed on two cards. In the NVIDIA Control Panel have a special parameter OpenGL rendering GPU, where you can choose which GPU will be responsible for rendering OpenGL (img01.png).
But on Linux (I have a CentOS 6.6 Linux with the driver version 340.58), the system sees two GPU, and the second uses as a co-accelerator in the calculations using NVIDIA CUDA or OpenCL, but at the same time, OpenGL is processed by only one GPU - Quadro K2000.
How can I configure the driver, in order to allow for OpenGL uses two GPU? Similar to the embodiment with Windows.
P.S. I tried to use
% nvidia-xconfig --multigpu=on
but it did not give the desired result, because the application shuts down during the start phase. And to be precise, at the stage of creating viewports and definition graphics accelerators.