470.14 - WSL with W10 Build 21343 - NVIDIA-SMI error

Accelerated Computing CUDA CUDA on Windows Subsystem for Linux
1

Is nvidia-smi broken? I’m on an AMD build with 2060 Super. Windows Insider 21343 (updated today) and 470.14.

Ubuntu 20.04, CUDA toolkit 11.2 installed using the WSL shell script that does not install Linux drivers.

I copied nvidia-sma to /usr/bin as instructed. When run, I get:

NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Failed to properly shut down NVML: Driver Not Loaded

10 Likes
2

Yeah, it’s the same with me.
Anyone can help?

3 Likes
3

I have the same issue. Ubuntu 20.04, CUDA 11.2, Windows 21343.1000, 5.4.72-microsoft-standard-WSL2 kernel. I’ve tried reinstalling CUDA.

4 Likes
4

Yep, same issue.

  • RTX 3060, drivers 470.14
  • Windows Insider 21343
  • Ubuntu 18.04 LTS,
  • Tried CUDA Toolkit 11.1.1 and CUDA Toolkit 11.2
    I followed this tutorial: CUDA on WSL :: CUDA Toolkit Documentation, BlackScholes sample works fine - no errors, but
    nvidia-smi or sudo docker run --rm --gpus all nvidia/cuda:11.1-base nvidia-smi returns:
    Failed to properly shut down NVML: Driver Not Loaded
    Reinstalling the drivers, Ubuntu, CUDA Toolkits also didn’t work.
5 Likes
5

Similar issue here.

  • Windows 10 Build 21354.1
  • CUDA Toolkit 11.2
  • Ubuntu 18.04 LTS, 5.4.72-microsoft-standard-WSL2
  • GTX 1050Ti, driver 470.14
1 Like
6

Same here.
Windows 10 Pro 21359.1
Ubuntu 18.04
WSL kernel 5.10.16
CUDA Toolkit 1.3.0 (also tried 11.0 and 11.2)
GTX 950, driver 470.14

Created this issue: https://developer.nvidia.com/nvidia_bug/3296440

2 Likes
7

Still there on newest insider build

1 Like
8

Same problem here with the latest attempts on 24Apr21 with latest Insider Build…
Windows 10 Pro 21364
Ubutnu 18.04.5 LTS
Linux version 5.10.16.3-microsoft-standard-WSL2 (oe-user@oe-host)
Cuda Toolkit 11.3
RTX 3070, driver 470.14 (Also updated via Windows to 470.25)…

I’d appreciate if Nvidia could respond to this!

4 Likes
9

Additionally, I have tried to follow CUDA on WSL :: CUDA Toolkit Documentation exactly…the cuda samples work but none of the Docker Container examples work.

6 Likes
10

ah, i already have multiple hours into this. Good to see that i’m not the only one with this problem.
I’m on windows 21364
nvidia-smi (powershel) 470.14 CUDA 11.3 (geforce 2080)
Tried Ubuntu 18.04 and 20.04

4 Likes
11

Same here:
Windows 21364.100
NVIDIA Driver 470.25 CUDA 11.4 (GeForce GTX 1050 Ti)
Ubuntu 20.04.2 LTS

EDIT: Tried Tensorflow GPU test and that works as well as the Blackscholes sample.

1 Like
12

Same here

RTX 2080, drivers 470.14
Windows Insider 21364
Ubuntu 18.04

emulcahy@DESKTOP-5C1NA5P:/c/Users/Eogha$ nvidia-smi

NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Failed to properly shut down NVML: Driver Not Loaded

1 Like
13

Same on the windows build 21370
wsl2 ubuntu 20.04.2 lts
gtx1080

docker: Error response from daemon: OCI runtime create failed: container_linux.go:367: starting container process caused
: process_linux.go:495: container init caused: Running hook #1:: error running hook: exit status 1, stdout: , stderr: nv
idia-container-cli: initialization error: driver error: failed to process request: unknown.

NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Failed to properly shut down NVML: Driver Not Loaded

2 Likes
14

Same problem.

RTX 2080
Driver: 470.14
Windows build: 21370.1
WSL2: 5.4.72-microsoft-standard-WSL2
Ubuntu 20.04.2 LTS

Followed the steps in CUDA on WSL User Guide, including section 7.3. Known Limitations point 2.

15

Finally some light on this issue. It’s a bug that will be fixed in an upcoming driver.

nvidia-docker 2.6.0-1 - not working on Ubuntu WSL2 · Issue #1496 · NVIDIA/nvidia-docker (github.com)

Docker Desktop still works because it has its own nvidia libraries.

5 Likes
16

my wsl2 details, that works without issues

  1. uname -a: 5.10.16.3-microsoft-standard-WSL2
  2. Driver Version: 470.14
  3. CUDA Version: 11.3
  4. OS build: 21370.1

And i execute nvidia-smi.exe not nvidia-smi

3 Likes
17

Rajkumarsaswamy:
That is Windows version of nvidia-smi. Not the wsl version.

1 Like
18

the same problem：
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Failed to properly shut down NVML: Driver Not Loaded

how can I fix it？

19

NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Running WSL on my Windows 10 machine (not pro). What is the issue?

20

I reinstall the wsl2 driver 465.21 then everything works well as original.
Given the version of 470.14, I think the solution is to find out the meta-package nvidia-utils-470.