RGM: Rust GPU Monitor

A lightweight GPU monitoring utility built with Rust and egui. Supports NVIDIA (via NVML) and AMD (via sysfs/hwmon, including integrated GPUs). Simple, fast, and reliable.

Features

• Multi-Vendor: Automatically detects and monitors NVIDIA or AMD GPUs.
• iGPU Friendly: Works with AMD integrated GPUs – unavailable sensors gracefully fall back to zero.
• Real-time Plots: GPU utilization, memory, temperature, and power visualised over time.
• Per-Process VRAM (NVIDIA): Which processes are holding your video memory.
• Throttle Reporting (NVIDIA): Why clocks are being held back, and how much of your load time the power limit has cost you so far.
• Low Overhead: No daemon, no root, a single binary.
• Desktop Integration: .deb/.rpm packages install an application entry in app launchers (Show Apps).

Prerequisites

Before you begin, ensure you have the following installed on your system:

1. Rust & Cargo: If you don't have them, install them from rust-lang.org.
2. GPU Drivers (one of the following):
   • NVIDIA: Official NVIDIA drivers installed. Verify with nvidia-smi.
   • AMD: The amdgpu kernel driver (included in most Linux kernels). Verify with ls /sys/class/drm/card*/device/driver pointing to amdgpu.

Installation

Method 1: Download Binary (Recommended)

Download the latest release for your Linux distribution from the Releases Page.

• Debian/Ubuntu: Download the .deb file and install:

  sudo dpkg -i rgm-ui_*.deb

• Fedora/RHEL/openSUSE: Download the .rpm file and install:

  sudo rpm -U rgm_ui-*.rpm

• Other Linux: Download the .tar.gz and extract it; the binary is at rgm/rgm inside the archive. Run it from there or copy it into a directory on your PATH.

Method 2: Install from Crates.io

If you have Rust installed, you can install RGM directly from crates.io:

cargo install rgm_ui

Method 3: Build from Source

1. Clone the repository:

   git clone https://github.com/Neomelt/RGM.git
   cd RGM

2. Build and install:

   cargo install --path .

Usage

The .deb/.rpm packages and both cargo install methods put the rgm binary on your PATH (the packages also add an app-launcher entry):

rgm

For the portable .tar.gz, run rgm/rgm from the extracted directory instead.

For development from a source checkout, use cargo run --release instead.

The application will auto-detect your GPU vendor and display real-time metrics.

Two flags answer without opening a window, which is what you want from a script:

rgm --version
rgm --help

Updating

Releases are not delivered by a package repository yet, so an update is a download plus an install. This checks the signature of what it downloaded and tells you which version you are on first:

rgm --version
cd /tmp && gh release download --repo Neomelt/RGM \
    --pattern '*_amd64.deb' --pattern 'SHA256SUMS' --clobber \
  && sha256sum -c --ignore-missing SHA256SUMS \
  && sudo dpkg -i rgm-ui_*_amd64.deb

Swap the pattern for *.rpm on Fedora/RHEL/openSUSE and dpkg -i for rpm -U. With Rust installed, cargo install rgm_ui upgrades in place.

---

Troubleshooting

NVIDIA

Error: libnvidia-ml.so: cannot open shared object file

This occurs when the application cannot find the NVIDIA Management Library (NVML), even if nvidia-smi works correctly.

Solution:

1. Find the NVML library path.

   ldconfig -p | grep libnvidia-ml.so.1

2. Create a symbolic link.

   sudo ln -s /lib/x86_64-linux-gnu/libnvidia-ml.so.1 /lib/x86_64-linux-gnu/libnvidia-ml.so

AMD

Missing metrics (VRAM, fan speed, etc.)

Some AMD integrated GPUs do not expose all sysfs nodes. This is normal – RGM will display 0 for any unavailable metrics.

Permission denied reading sysfs

Running as a regular user should be sufficient for read-only monitoring. If you see permission errors, ensure your user has read access to /sys/class/drm/card*/device/.

License

This project is licensed under either of:

• Apache License, Version 2.0, (LICENSE-APACHE or https://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or https://opensource.org/licenses/MIT)

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this project by you, as defined in the Apache-2.0 license, shall be dually licensed as above, without any additional terms or conditions.