Gfilui

I'm trying to make Cuda work with a second GPU that I have recently added to my workstation rig.

When I used to download and install Cuda in my computer, I had the Nvidia NVS 315 GPU. Now, I've added a Quadro P400 GPU, and I'm looking to install the latest Cuda version for it to leverage it for accelerated computing while running Deep Learning algorithms (Caffe2 for instance). However, I don't know what to start with.

Check the following outputs :

nvidia-smi

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.111 Driver Version: 384.111 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 NVS 315 Off | 00000000:01:00.0 N/A | N/A |
| 30% 45C P0 N/A / N/A | 255MiB / 956MiB | N/A Default |
+-------------------------------+----------------------+----------------------+
| 1 Quadro P400 Off | 00000000:05:00.0 Off | N/A |
| 34% 34C P8 ERR! / N/A | 1MiB / 1999MiB | 0% Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 Not Supported |
+-----------------------------------------------------------------------------+

nvcc --version

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2015 NVIDIA Corporation
Built on Tue_Aug_11_14:27:32_CDT_2015
Cuda compilation tools, release 7.5, V7.5.17

which nvcc

/usr/bin/nvcc

ldconfig -p | grep cuda

libnvToolsExt.so.1 (libc6,x86-64) => /usr/local/cuda-9.0/targets/x86_64-linux/lib/libnvToolsExt.so.1
libnvToolsExt.so (libc6,x86-64) => /usr/local/cuda-9.0/targets/x86_64-linux/lib/libnvToolsExt.so
libicudata.so.55 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libicudata.so.55
libicudata.so.55 (libc6) => /usr/lib/i386-linux-gnu/libicudata.so.55
libicudata.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libicudata.so
libcudart.so.9.0 (libc6,x86-64) => /usr/local/cuda-9.0/targets/x86_64-linux/lib/libcudart.so.9.0
libcudart.so.7.5 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libcudart.so.7.5
libcudart.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libcudart.so
libcuda.so.1 (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libcuda.so.1
libcuda.so.1 (libc6) => /usr/lib/i386-linux-gnu/libcuda.so.1
libcuda.so (libc6,x86-64) => /usr/lib/x86_64-linux-gnu/libcuda.so
libcuda.so (libc6) => /usr/lib/i386-linux-gnu/libcuda.so

I want to make my algorithms select the second GPU (Quadro P400) for accelerated computing

To answer my question, yes it's a mess (it was).

I completely removed Cuda by following the instructions found here (uninstalling cuda with its dependencies), then I installed Cuda 9.0 by following the instructions contained here (you can skip the uninstallation part if you did it properly before).

After that, in order to install the Cuda libraries within a virtual environment set by Conda and install within the Caffe2 library, I followed the following instructions:

• Setting up of a virtual env


• Installing Cuda libraries


• Installing the Caffe2 library compatible with Cuda 9.0

conda install -c caffe2 caffe2-cuda9.0-cudnn7