NVIDIA SMI Command Examples
Understanding NVIDIA SMI
NVIDIA System Management Interface (nvidia-smi
) is a command-line utility that provides a way to monitor and manage NVIDIA GPU devices. It is an essential tool for system administrators, developers, and researchers working with NVIDIA hardware, especially in environments involving high-performance computing, deep learning, and AI. This tool allows you to query GPU status, such as utilization, memory usage, temperature, power consumption, and driver versions.
Common NVIDIA SMI Commands
Listing Available NVIDIA Devices
To see a list of all NVIDIA GPUs detected by the system, use the -L
flag:
nvidia-smi -L
Querying Specific GPU Details
You can query specific details about each GPU and format the output as CSV for easier parsing. This command lists the index, name, UUID, and serial number of each GPU:
nvidia-smi --query-gpu=index,name,uuid,serial --format=csv
Detailed GPU Information
To get comprehensive information about a specific GPU (e.g., GPU with index 0), use the -q
flag:
nvidia-smi -i 0 -q
System Topology Information
Understanding how GPUs are connected to the CPU and other devices is crucial for performance optimization. The topo
command displays the system topology in a matrix format:
nvidia-smi topo --matrix