nvtop

nvtop is a terminal monitor that shows activity across multiple GPUs and accelerators in an interface familiar to htop users. It's useful for people running local LLM workloads or managing compute servers who want to see which processes are using their hardware and how much GPU memory they consume. It runs locally on Linux and is open source under GPLv3 or later.

The interface combines process information with charts for GPU and memory utilization, temperature, power draw, clock speeds and PCIe traffic. You can choose which metrics appear and save your preferences between sessions. Available measurements depend on the device and its driver, so support doesn't mean every card exposes the same information.

Hardware coverage extends beyond NVIDIA, AMD and Intel GPUs. nvtop also supports Qualcomm Adreno, Huawei Ascend, Iluvatar CoreX, MetaX and Enflame hardware, along with Broadcom VideoCore and Rockchip NPUs. Tenstorrent support covers Blackhole, Wormhole and Grayskull accelerators, with process listings and readings for temperature, power, clocks and fans. This breadth makes it relevant to machines with several kinds of compute hardware.

Linux support relies on the corresponding hardware drivers, including NVIDIA's NVML and Intel's i915 or Xe drivers. Docker is another supported way to run it, and NVIDIA monitoring works under WSL2 when the vendor's GPU support is available. Apple M1 and M2 support uses Metal, has limitations, and monitors only Apple hardware in the Apple build.