So I have just bought a Quadro K4000 (Kepler) and found out it does not support double precision (DP). I can get another Quadro 4000 (Fermi) and it does support DP.
Can I use them together where K4000 for graphics and 4000 for CUDA number crunching?? How do I assign them for these different purpose?
FEA softwares. These two cards are from different architecture, Fermi vs Kepler. Will they work together?
Thanks, tried it. But the rep seems clueless.
He said they will work together using unified driver, but I can't assign K4000 for graphics only and 4000 for CUDA compute only, there is no option he says.
Does your FEA software have the relevant support for CUDA processing? Reason I am asking is that I have had experience with FEA software requiring an additional HPC license for GPU compute…
You might also be able to check the FEA software's configuration - you may be able to target the CUDA GPU there.
You also need to look at it from the perspective that you can buy headless nVidia GPU's specifically for CUDA compute.. So there has to be a way to target it somehow..
Sorry, that's all I can really add to the discussion.
What software are you using?
If I recall you can set the cuda processor in the nvidia control panel.