$ cd pranavchip.com && cd /docs -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512 Tuning Nvidia GPU for max performance ===================================== Pranav Chiploonkar2026-07-08 1. Enable NVIDIA Persistence Daemon =================================== This keeps the GPU driver loaded and ready, even when idle. This improves snappiness when an application requires GPU time, because the kernel hasn't unloaded it. This is a simple fix, because Nvidia provides a systemd service for precisely this: $ sudo systemctl enable --now nvidia-persistenced.service You can check its status via: $ systemctl status nvidia-persistenced.service ● nvidia-persistenced.service - NVIDIA Persistence Daemon Loaded: loaded (/usr/lib/systemd/system/nvidia-persistenced.service; enabled; preset: disabled) Active: active (running) since Wed 2026-07-08 19:47:27 MST; 45min ago Invocation: 1cac2c7019fe4ddcb052002875a9bf75 Main PID: 35474 (nvidia-persiste) Tasks: 1 (limit: 18462) Memory: 764K (peak: 2.1M) CPU: 9ms CGroup: /system.slice/nvidia-persistenced.service └─35474 /usr/bin/nvidia-persistenced --user nvidia-persistenced Jul 08 19:47:27 candypaint systemd[1]: Starting NVIDIA Persistence Daemon... Jul 08 19:47:27 candypaint nvidia-persistenced[35474]: Started (35474) Jul 08 19:47:27 candypaint systemd[1]: Started NVIDIA Persistence Daemon. 2. Ensure power limit is set to its maximum =========================================== I have an NVIDIA RTX 2020 Mobile, and its max power is 80 watts. It is already set to 80 W, but just in case, it is possible to enforce it via: $ sudo nvidia-smi -pl 80 Changing power management limit is not supported for GPU: 00000000:01:00.0. Treating as warning and moving on. All done. Uh, so maybe not on my card in particular, but it's useful to know about regardless? 3. Set GPU clocks and VRAM clocks to max ======================================== The max clock speeds can be found in $ sudo nvidia-smi -q Upon finding them, you can set the min/max to be the same. This will pin clock speeds to the max at all times, even while idling. Note that this will *significantly* increase thermal load at idle. $ sudo nvidia-smi --lock-gpu-clocks=2100,2100 $ sudo nvidia-smi --lock-memory-clocks=5501,5501 To undo this, run $ sudo nvidia-smi --reset-gpu-clocks $ sudo nvidia-smi --reset-memory-clocks -----BEGIN PGP SIGNATURE----- iHUEARYKAB0WIQSJp6PBWSJCR8aCbbXYWutBuYPnewUCak8Y/gAKCRDYWutBuYPn e+pXAP49HzX9JMvuO4Vy5fdKCFnkURMiiiDG1wbLKbR9s8cI+gEAmwfvm01Hibsy JKkNDJZK7W/YhA0Dy9ulgfwu5pVGNwE= =YWXq -----END PGP SIGNATURE-----