Hi all,
I have recently installed both tensorflow-gpu, nvidia drivers and cuda toolkit. After some research I determined I had to install CUDA 9.1 in order to be compatible with my nvidia driver which is 390.77 (I have a GTX 1070) [source: https://devtalk.nvidia.com/default/topic/1038069/cuda-setup-and-installation/which-cuda-version-will-work-with-driver-390-77-/]. However I still get the following error message:
“CUDA driver version is insufficient for CUDA runtime version”
- Output from nvidia-smi:
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.77 Driver Version: 390.77 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 1070 Off | 00000000:01:00.0 On | N/A |
| 0% 46C P8 13W / 200W | 473MiB / 8105MiB | 1% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1115 G /usr/lib/xorg/Xorg 24MiB |
| 0 1150 G /usr/bin/gnome-shell 51MiB |
| 0 1401 G /usr/lib/xorg/Xorg 209MiB |
| 0 1533 G /usr/bin/gnome-shell 139MiB |
| 0 1986 G ...uest-channel-token=16350887306611312583 44MiB |
+-----------------------------------------------------------------------------+
- Output from nvcc -V:
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85
- Output from lspci
01:00.0 VGA compatible controller: NVIDIA Corporation GP104 [GeForce GTX 1070] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GP104 High Definition Audio Controller (rev a1)
What am I doing wrong? Any help is much appreciated!
Kind regards
EDIT
I did some more digging and found that the cudatoolkit specified in conda was 9.2. The installation of 9.1 didn’t seem to work so I settled on 9.0 (did the complete reinstall using
conda install tensorflow-gpu==1.11 cudatoolkit==9.0 cudnn==7.1.2 h5py
).
So now it is working but i would still like soma advice that says if this is a good solution or that I am now running things sub-optimally because of the lower cuda version.
Thanks in advance!