skilld
All skills
nvidia avatar

/jetson-customize-clocks

@87053a0
by NVIDIA Corporationnvidia/skills3.5k stars
424

Use to lock/cap Jetson CPU/GPU/EMC clocks, toggle EMC/CPU DVFS, or change cpufreq governors by editing BPMP DTB and nvpower.sh pre-flash. Do NOT use for live tuning or nvpmodel edits.

Requires /jetson-customize-nvpmodel /jetson-set-target /jetson-init-target /jetson-init-image /jetson-init-source /jetson-promote-image /jetson-flash-image

Ask your Agent

Use this Skill: https://skilld.dev/gh/nvidia/skills/jetson-customize-clocks

This session only. Nothing lands on disk.

referencesnvpower-sh-edits.md

≈477 tokens on demand. Your agent reads this file only when SKILL.md points to it.

nvpower.sh edits

Reference material for ../SKILL.md Operation 2.

nvpower.sh lives at Linux_for_Tegra/rootfs/etc/systemd/nvpower.sh and runs at boot via nvpower.service. It sets cpufreq / devfreq governors and (optionally) per-device min / max / static rates.

Functions of interest

Function Writes to Default (non-safety platforms)
set_cpufreq_governor() /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor schedutil
set_devfreq_governor() /sys/class/devfreq/<dev>/governor tegra_wmark — but GPU and nvjpg are skipped

There is no dedicated function for per-device min / max / static rates. To pin or bound a device's rate, write min_freq / max_freq to its sysfs node from a helper added inside (or invoked alongside) the existing flow.

Common edits

Goal Edit
Pin all CPUs to Fmax In set_cpufreq_governor: set desired_cpufreq_gov="performance" unconditionally (remove the IS_SAFETY_PLATFORM → schedutil override).
Pin all devfreq devices to Fmax In set_devfreq_governor: set desired_devfreq_gov="performance" and remove the gpu / nvjpg skip block.
Static custom rate on one device After the governor is set, write <rate> to min_freq then max_freq (order matters when raising; reverse when lowering).
Min / max bounds without pinning Same as static, but min < max — governor stays dynamic within the bounds.

Package-upgrade caveat

nvpower.sh is shipped by the nvidia-l4t-init deb. Package upgrades clobber hand edits to the rootfs file. For long-lived test setups, prefer a systemd drop-in or a separate helper file checked into the rootfs alongside nvpower.sh rather than editing it in place.

Apply

nvpower.service runs the edited script on the next boot after Deploy re-flashes the rootfs — see the sibling skill /jetson-flash-image.

Source: SKILL.md on GitHub

Third-party checks

2 warnings3mo3 checks · Risk MEDIUM
  • Gen Agent Trust Hub3mo

    The skill facilitates NVIDIA Jetson clock and power customization. It involves administrative tasks using sudo and processes external system configuration files, creating a potential surface for indirect prompt injection.

  • Socket3mo

    No alerts

  • Snyk3mo

    Risk: MEDIUM · 1 issue

Provenance

Signed by skilld at 87053a0. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 3 months ago

Capability

version
0.0.1
Other metadata
metadata
{
  "data-classification": "public",
  "author": "Jetson Team",
  "tags": [
    "clocks",
    "cpu",
    "gpu",
    "emc",
    "dvfs",
    "bwmgr",
    "bpmp",
    "nvpower",
    "cpufreq",
    "devfreq"
  ],
  "domain": "clocks"
}

README badge

README badge for nvidia/skills/jetson-customize-clocks