How to Check GPU Usage in a Linux Server

When running machine learning models, video rendering, or high-performance computing tasks on a Linux server, monitoring GPU usage is essential. It helps ensure your GPU resources are efficiently used and prevents performance bottlenecks.

In this guide, you’ll learn how to check GPU utilization, temperature, memory usage, and running processes across different GPU types — NVIDIA, AMD, and Intel.

---

1. Check GPU Usage for NVIDIA GPUs

If you’re using an NVIDIA GPU, the most common tool is nvidia-smi, which comes with the NVIDIA driver package.

Command:

nvidia-smi

    

Output Example:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 545.23.06    Driver Version: 545.23.06    CUDA Version: 12.3     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        | Usage (%)  | Memory-Usage | Temp  | Fan | Processes      |
|------------------+-------------+--------------+-------+-----+---------------|
| 0  RTX 3090      | 78%         | 9500MiB / 24576MiB | 70C  | 45% | 4        |
+-----------------------------------------------------------------------------+

Common flags:

• nvidia-smi -l 1 → Live refresh every second
• nvidia-smi dmon → Detailed GPU monitoring
• watch -n 1 nvidia-smi → Watch mode for real-time updates

---

2. Check GPU Usage for AMD GPUs

For AMD GPUs, use the radeontop command.

Install:

sudo apt install radeontop

Run:

sudo radeontop

It displays a real-time console-based dashboard with GPU load, memory usage, and temperature.

Alternatively, you can check information using:

cat /sys/class/drm/card0/device/gpu_busy_percent

---

3. Check GPU Usage for Intel GPUs

For Intel integrated GPUs, install the intel-gpu-tools package.

Install:

sudo apt install intel-gpu-tools

Run:

sudo intel_gpu_top

You’ll get an ncurses-based live GPU usage graph, similar to top or htop for CPU.

---

4. Using Generic Monitoring Tools

If you’re managing multiple servers or mixed hardware, you can use tools like:

• glances — a cross-platform system monitor with GPU support.
• gpustat — a simple utility to check NVIDIA GPU usage in a compact format:
  

  

  pip install gpustat
  gpustat
  

• nvtop — a “top”-like interface for NVIDIA GPUs:

  sudo apt install nvtop
  

---

Summary

GPU Type	Tool	Command
NVIDIA	nvidia-smi	nvidia-smi
AMD	radeontop	sudo radeontop
Intel	intel-gpu-tools	sudo intel_gpu_top
All	glances, gpustat, nvtop	Various

Monitoring GPU performance ensures that your server resources are used optimally and helps detect memory or thermal issues early.

---

 

How to Check GPU Usage in a Linux Server (F.A.Q)