nv-monitor now supports Jetson Orin (Nano, NX, AGX)
We’ve added full GPU monitoring support for Jetson devices in nv-monitor (GitHub - wentbackward/nv-monitor: A lightweight terminal system monitor built for the **NVIDIA DGX Spark** (Grace CPU + GB10 GPU). · GitHub) — a lightweight terminal monitor, CSV logger, and Prometheus exporter for NVIDIA GPU systems, all in a single <80KB binary.
Now you can simply get all your edge devices into grafana with no dependecies and minimal system overhead
What’s new for Jetson:
- GPU utilization via Tegra sysfs (auto-detected across JetPack versions)
- GPU temperature from thermal zones
- ARM Cortex-A78AE core type labels
- Legacy arm64 binary (glibc 2.35) available — no need to build from source
- Unified memory handled gracefully
Performance on Orin Nano Super (8 GB, 6 cores):
| nv-monitor | top | |
|---|---|---|
| Binary | 64 KB | 118 KB |
| CPU usage | 0.20% | 2.10% |
| Monitors GPU | Yes | No |
| Prometheus export | Yes | No |
10x less CPU than top, while monitoring CPU, memory, GPU, processes, and optionally exporting Prometheus metrics for Grafana dashboards across a fleet.
Works on DGX Spark, Jetson Orin family, and any Linux system with an NVIDIA GPU. Also includes demo-load, a zero-dependency synthetic load generator for validating your monitoring pipeline end-to-end.
System load is generated by the included demo-load sinusoidal synthetic system load generator. This gives more interesting data than similar tools, which just max out all resources.
Links:
- GitHub: GitHub - wentbackward/nv-monitor: A lightweight terminal system monitor built for the **NVIDIA DGX Spark** (Grace CPU + GB10 GPU). · GitHub
- Download: Release v1.7.0 · wentbackward/nv-monitor · GitHub
- Performance report: nv-monitor/PERFORMANCE.md at main · wentbackward/nv-monitor · GitHub
If you find this useful, a star on the repo would be much appreciated!