I do a lot with LLM (and now especially Local LLM on my GPU) so I use that nvtop by nVidia quite a bit to see what’s going on with my card. Ergo “we” (Claude and I) whipped up this last night too.
Images might be a little outdated, but this gives you a good idea…
vobot-apps/nvtop at main · DAE51D/vobot-apps · GitHub
vobot-apps/nvtop-daemon at main · DAE51D/vobot-apps · GitHub
in case you don’t want to wait for gallery approval
NOTICE: this REQUIRES a small lightweight daemon service running on your linux host where the GPU is. This sends back some JSON that the Vobot app then reads. Sorry, I don’t have an OSX or Windows daemon (yet) as my GPU is in my Proxmox host and runs the Local LLM LXCs on it (shared GPU of course). But PLEASE feel free to send me a PR if you build one.
curl ``http://proxmox.home.lan:8039/api/gpu-data`` | jq .
{
"gpus": {
"0": {
"index": "0",
"name": "NVIDIA GeForce RTX 5060 Ti",
"uuid": "GPU-2bb45349-6461-e802-99c1-098bae63ad20",
"driver_version": "595.71.05",
"temperature": 33.0,
"utilization": 0.0,
"memory_utilization": 30.0,
"memory_used": 5478.0,
"memory_total": 16311.0,
"memory_free": 10365.0,
"power_draw": 4.64,
"power_limit": 180.0,
"fan_speed": 0.0,
"clock_graphics": 210.0,
"clock_sm": 210.0,
"clock_memory": 405.0,
"clock_max_graphics": 3120.0,
"clock_max_sm": 3120.0,
"clock_max_memory": 14001.0,
"pcie_gen": "1",
"pcie_gen_max": "3",
"pcie_width": "8",
"pcie_width_max": "16",
"performance_state": "P8",
"throttle_reasons": "None",
"timestamp": "2026-07-20T15:15:07"
}
},
"processes": {
"0": [
{
"pid": "3237629",
"process_name": "/opt/llama-cpp/bin/llama-server",
"used_memory": 5460.0,
"cpu_percent": 0.5,
"rss_kib": 3487724,
"command": "/opt/llama-cpp/bin/llama-server --cache-reuse 256 --chat-template-kwargs {\"enable_thinking\": false, \"preserve_thinking\": false} --host 127.0.0.1 --metrics --min-p 0.05 --no-mmap --port 35329 --presence-penalty 0.0 --reasoning-budget 0 --repeat-penalty 1.05 --sse-ping-interval 15 --temperature 0.2 --tools all --top-k 40 --top-p 0.9 --webui-config-file /root/systemd/llama-ui-mcp-config.json --webui-mcp-proxy --no-warmup --alias ornith-9b-q4km --batch-size 512 --ctx-size 32768 --cache-type-k q4_0 --cache-type-v q4_0 --flash-attn on --model /mnt/media/gguf/ornith-9b/ornith-1.0-9b-Q4_K_M.gguf --n-predict 4096 --n-gpu-layers 99 --parallel 1 --reasoning off --ubatch-size 128"
}
]
},
"timestamp": 1784585707.8400185,
"git_commit": "ddb5741"
}