How to optimize the cold boot time of Xavier NX

Jetson Xavier NX Boot Time Discrepancy - Help Request

Hardware and Setup

  • Board: Jetson Xavier NX Developer Kit

  • Flash method: NVIDIA SDK Manager

  • Boot device: NVMe SSD

  • JetPack version: 5.1.6

systemd-analyze Result

After flashing, systemd-analyze reports approximately 29 seconds:

  • Kernel: 7.219s

  • Userspace: 22.430s

  • Total: 29.649s

  • graphical.target reached after 21.590s in userspace

Actual Measured Boot Time

Using a phone stopwatch from the moment power is applied until the desktop is fully visible: 1 minute 20 seconds (~80 seconds).

The Problem

There is a large gap (~50 seconds) between the systemd-reported boot time and the real-world boot time. I have spent considerable time searching NVIDIA Developer Forums and other resources but have not found a solution.

Request for Help

I am looking for guidance on why the measured boot time is much longer than systemd-analyze reports, and how to reduce the overall time from power-on to desktop.

Hi,

Please share the serial console log for further investigation.

Thanks

串口_2026-06-23_19_35_17.log (80.7 KB)
The uploaded file is the console log when Xavier boots up
The uploaded file is the console log when Xavier boots up

Hi,

The key point: systemd-analyze only measures from kernel start, so the ~50 s gap is in the firmware/UEFI stage, which it can’t see.

Thanks

Hi @zengwanq,

Yes, it is possible to reduce the overall boot time on Jetson Xavier NX.

One possible explanation is that systemd-analyze measures the time from when the Linux kernel starts until the specified systemd target is reached. It may not account for earlier stages such as BootROM, MB1/MB2, UEFI, or other firmware initialization, which could contribute significantly to the total time you observe from power-on to a usable desktop.

NVIDIA provides official boot-time optimization guides that describe techniques for reducing both bootloader and kernel boot times.

You can find them here:

If, after applying these optimizations, you still see a large discrepancy between systemd-analyze and the actual power-on-to-desktop time, sharing a serial boot log could help identify which boot stage is taking the extra time.

If you have any further questions, feel free to ask!

Regards,

Oscar Mendez
Embedded SW Engineer at RidgeRun
Contact us: support@ridgerun.com
Developers wiki: https://developer.ridgerun.com/
Website: www.ridgerun.com