Can DLS utilize Maximus double card GPU system?


#1

Hello,

I’m working on a DELL workstation with a double card configuration -called by Nvidia a “Maximus” system. The two cards are a Quadro 4000 and a Tesla C2075. I was very sad to find out that Deep Learning Studio (Desktop edition) cannot use the particular GPUs. The log is saying:
“Checking GPU support…
GPU not supported”

I understand that the compatibility declared in the website is for GPUs with “compute capability > 3.0”. But on the other hand a Maximus system is intended to be used in exactly demanding computational situations (especially for machine learning purposes).

Is there a way of making Deep Learning Studio to use the computing capabilities of this GPU configuration?

Michael Vinos


#2

DLS uses MxNet backend. MxNet uses CUDA interface and is compiled for selected CUDA architectures. So if MxNet does not support your GPU, DLS will also be no able to support it.

Have you tried running MxNet/Tensorflow code in a python environment ?