CUDA Runtime Init issue

Hi All,

I recently got a RTX 4000 Ada installed on a Rocky Linux 9.5 server.

The system sees the GPU and driver installation seems fine,

 [root@jupyter2 ~]# nvidia-smi
Tue Oct 21 12:19:09 2025
±----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.95.05              Driver Version: 580.95.05      CUDA Version: 13.0     |
±----------------------------------------±-----------------------±---------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA RTX 4000 SFF Ada …    On  |   00000000:41:00.0 Off |                  Off |
| 30%   36C    P8              6W /   70W |       1MiB /  20475MiB |      0%      Default |
|                                         |                        |                  N/A |
±----------------------------------------±-----------------------±---------------------+

±----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
±----------------------------------------------------------------------------------------+

But I’m having issues getting the GPU recognized when running simple tests.

Using the CUDA sample binaries fails

./deviceQuery
./deviceQuery Starting…

CUDA Device Query (Runtime API) version (CUDART static linking)

cudaGetDeviceCount returned 3
 → initialization error
Result = FAIL

Pytorch also fails to recognize the GPU. This is torch installed with pip in a conda environment

python -c “import torch; print(torch.version.cuda); print(torch.cuda.is_available())”
13.0
/home/henryz/miniconda3/envs/torch/lib/python3.11/site-packages/torch/cuda/init.py:182: UserWarning: CUDA initialization: CUDA driver initialization failed, you might not have a CUDA gpu. (Triggered internally at /pytorch/c10/cuda/CUDAFunctions.cpp:119.)
return torch._C._cuda_getDeviceCount() > 0
False

Any ideas on how to proceed?