Orin Nano Dev Kit (P3767-0005): boots only after SDK Manager flash, kernel panic on every subsequent reboot — RMA?

NVIDIA Customer Care directed me here (case 260713-000362).

Hardware: Jetson Orin Nano Developer Kit — module P3767-0005, carrier P3768-0000. Purchased June 2026 (Amazon; invoice available). Approximately 1 month old, within warranty.

Problem: The board boots successfully immediately after a fresh SDK Manager flash. The first subsequent reboot or shutdown results in a black screen and kernel panic (blinking keyboard lock LEDs), and it never boots again. Only another full SDK Manager firmware+OS reflash restores a single boot — then the next reboot fails identically.

JetPack versions tested: 6.2.1 (L4T 36.4.4) and 6.2.2 (L4T 36.5). Identical failure on both.

Troubleshooting completed:

  • SD card passed a full write/verify surface test (H2testw, 122 GB, zero errors)
  • SD image size verified against NVIDIA’s published byte count
  • An SD-only reflash (Etcher) does not recover the board — only SDK Manager, which also rewrites QSPI firmware, does
  • Fails with all peripherals disconnected (CSI camera, USB-C, ethernet)
  • Shutdown verified clean. (nv-l4t-usb-device-mode was observed ignoring SIGKILL and stalling an earlier shutdown; it was masked, and shutdowns are now clean — the boot still fails.)
  • Immediately before the final failed reboot: systemctl --failed reported 0 failed units; nvbootctrl dump-slots-info showed version 36.5.0, capsule update status 0, bootloader slot A, both slots “normal”
  • Auto-updates disabled and nvidia-l4t-* pinned, so no background firmware updates are occurring
  • Pressing Esc at power-on produces no UEFI menu on a failed boot
  • SSH to a static IP fails on failed boots, confirming the system is genuinely not booting (not merely a display/EDID issue)

Assessment: A clean factory flash cannot survive a single reboot, with the SD card and OS image both independently verified and all peripherals removed. This appears to be a hardware fault in persisting boot state across power cycles.

Question: Does this qualify for an RMA? I can provide the serial number and purchase invoice. Thank you.

— 🤖 This is an automated AI response. I’m here to help, but please verify important details! —

Based on the provided context, it appears that the Jetson Orin Nano Developer Kit is experiencing a hardware issue, where it boots successfully only after a fresh SDK Manager flash, but fails to boot on subsequent reboots, resulting in a kernel panic and black screen. This issue persists even after troubleshooting and verifying the SD card and OS image.

Given the information provided, including the fact that the issue occurs with multiple JetPack versions (6.2.1 and 6.2.2) and that the board is within warranty, it seems that the problem may be related to the hardware’s ability to persist boot state across power cycles.

The context does not provide a clear answer to whether this qualifies for an RMA (Return Merchandise Authorization). However, it does mention that NVIDIA Customer Care directed the user to the forum, suggesting that they may be aware of the issue and potentially willing to assist further.

To determine the next steps, it would be best to follow up with NVIDIA Customer Care or a support engineer, providing them with the serial number, purchase invoice, and detailed description of the issue, as you have done. They can assess the situation and provide guidance on whether the issue qualifies for an RMA or if there are other troubleshooting steps that can be taken.

In summary, while the provided context supports the existence of a hardware-related issue affecting the Jetson Orin Nano’s ability to boot consistently, it does not explicitly state what actions NVIDIA will take in response to this specific problem. Further communication with NVIDIA support is necessary to resolve the issue.

— We support team will also look into your issue to have further suggestions soon. please stay tuned. —

Hi,

Please try to flash with the latest version JP7.2 and record the serial console log.

Thanks