Getting the error: 'cuda driver version is insufficient' - tried all solutions I could find online

Accelerated Computing CUDA CUDA Setup and Installation
1

I am attempting to run the TensorFlow GPU version but get the error when ever I create a TensorFlow Session.

OS: Linux 4.14
Distro: Manjaro 17.1.1
Machine: Dell XPS 15 9560
GPU: Nvidia GeForce GTX 1050
Drivers: ‘video-hybrid-intel-nvidia-bumblebee’
CUDA: 9.1
cuDNN: 7.0
TensorFlow: r1.5 built from source (which appears to be working fine for others: https://github.com/tensorflow/tensorflow/issues/15656)

optirun nvidia-smi
                                                                                                                                                                                             
Thu Jan 11 21:57:15 2018       
+-----------------------------------------------------------------------------+
| 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  GeForce GTX 1050    Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   38C    P0    N/A /  N/A |      6MiB /  4041MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      2024      G   /usr/lib/xorg-server/Xorg                      6MiB |
+-----------------------------------------------------------------------------+

optirun python test.py                                                                                                                                                                                         

2018-01-11 22:12:48.275062: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:898] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-01-11 22:12:48.275380: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1206] Found device 0 with properties: 
name: GeForce GTX 1050 major: 6 minor: 1 memoryClockRate(GHz): 1.493
pciBusID: 0000:01:00.0
totalMemory: 3.95GiB freeMemory: 3.89GiB
2018-01-11 22:12:48.275410: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1300] Adding visible gpu device 0
2018-01-11 22:12:48.275560: E tensorflow/core/common_runtime/direct_session.cc:168] Internal: cudaGetDevice() failed. Status: CUDA driver version is insufficient for CUDA runtime version
2

which cuda version did you install?

did you validate the CUDA install?

BTW manjaro linux is not an officially supported distro for CUDA

3

Sorry, CUDA 9.1, cuDDN 7.0, TensorFlow r1.5 built from source - which appears to be working fine for others: [url]https://github.com/tensorflow/tensorflow/issues/15656[/url]

How do I validate the CUDA installation? It has been installed to /opt/cuda/ and I have adjusted my ~/.bashrc accordingly.

4

CUDA 9.1 won’t work with a r384 driver. That is why you are getting that error message.

Get your installers from here: [url]http://www.nvidia.com/getcuda[/url]

Read the linux install guide. It explains validation and also how to install.

Be advised - installing in an optimus/bumblebee environment is non-trivial. Read the install guide carefully.