Setting Up GPU Docker Support: Understanding NVIDIA-SMI Output

When setting up GPU support for Docker, it’s crucial to ensure that your host system is compatible. One of the tools that can help you with this is NVIDIA System Management Interface (nvidia-smi), a command-line tool that reports detailed information about the current state of your NVIDIA GPU.

PS C:\Windows\system32> nvidia-smi
Fri Sep 22 08:52:17 2023
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 529.08       Driver Version: 529.08       CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  Quadro M1200       WDDM  | 00000000:01:00.0  On |                  N/A |
| N/A    0C    P8    N/A /  N/A |    763MiB /  4096MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
...

Key Information for Docker Compatibility

1. NVIDIA-SMI and Driver Version: The output shows that the NVIDIA-SMI version is 529.08, the driver version is also 529.08, and the CUDA version is 12.0. These versions are important to ensure compatibility with your CUDA applications and Docker images.
2. GPU Details: The GPU in use is a Quadro M1200. This information is crucial as different GPUs may support different features, and this could affect the compatibility with certain Docker images.
3. Performance Metrics: The fan speed, temperature, power usage, memory usage, and GPU utilization are displayed for each GPU. In this case, the Quadro M1200 is running at 0C with 763MiB of its 4096MiB memory in use, and it’s not currently being utilized (0%). These metrics can help you monitor your system’s performance when running GPU-accelerated Docker containers.
4. Running Processes: The last section lists all processes currently utilizing the GPU along with their PID, type (C+G for Compute and Graphics), process name, and GPU memory usage. This information can be useful for diagnosing issues related to resource contention when running multiple containers.

By understanding the output of nvidia-smi, you can ensure that your host system meets the requirements for running GPU-accelerated Docker containers and troubleshoot any issues that arise during setup or operation.