NVIDIA just published the official JetPack 7.2 release blog post titled “Deploy Agentic-Ready AI at the Edge with Memory Efficiency in NVIDIA JetPack 7.2” (Jetpack7.2). This release is significant for a few reasons directly relevant to Seeed hardware users:
1. Jetson Orin NX is now officially supported — JetPack 7.2 extends the Ubuntu 24.04 + kernel 6.8 + CUDA Toolkit 13.0 compute stack (first introduced with Jetson Thor) to the entire Jetson Orin family, including Orin NX. This means JetPack 7.2 is the first version in the JetPack 7.x line to support our Orin NX modules — making it highly relevant for J401 and J501 carrier board users.
2. Super Mode for Jetson AGX Orin 32 GB — JetPack 7.2 introduces Super Mode for the AGX Orin 32 GB, boosting GPU frequency from 930 MHz to 1.3 GHz and increasing AI performance from 200 TOPS to 241 TOPS (a 20%+ uplift), while reducing cost by 45% compared to the 64 GB module. This is a major value proposition for anyone using the AGX Orin 32 GB.
3. Other notable features: one-command NemoClaw deployment, NVIDIA Agent Skills for Jetson (memory optimization, Linux customization, model benchmarking), official Yocto Project support, and MIG on Jetson Thor.
Given all this, we’re wondering: does Seeed have any planned timeline to release official JetPack 7.2 images for the J401 and J501 carrier boards?
This is especially pressing given that many of us are still waiting on an official Seeed image for JetPack 6.2.2 — and now JetPack 7.2 with full Orin NX support is already out. The community understands Seeed needs time for rigorous testing and hardware-specific BSP work (device tree, power configs, etc.), but even a rough timeline or acknowledgment of the roadmap would be very helpful for planning our projects.
Hi community, I installed the new firmware, ubuntu 24.04 and the jetpack 7.2 using Nvidia’s official upgrade path and their image. Is there any reason to wait for Seeedstudio with their own release? Is there any specific BSP that won’t work with vanilla nvidia update?
SO , A couple things. 1. Seed can’t possibly know exactly when the newest Jetpacks will be released outside of what are the official release dates, then some testing needs to be performed and updates to the seeed specific differences in the reference designs and the REAL, new and different chips and some drivers as well , case in point would be the HDMI, from packs ago.
They have explained it on here, how the process works.
Nvidia has more of a “road-map” more so, than a release schedule if you are familiar with the cycles. I’m confident Seeed will get there, but So many AI , related products recently that require greater explanation than selling “grove Sensors”
,The ROBOTICS alone would keep a small team busy for hours.
they are adding more AI personnel too, so be patient and stay tuned.
AFAIK, the existing JP,2.2.2 update is working well? there is NOT an official JP6.2.2 it was always a APT -update.. Ymmv?
I.e.
Seeed Studio does not offer a direct, standalone system image specifically for JetPack 6.2.2. Instead, for Jetson devices like the reComputer J401, J501, and Orin Nano/NX, you should use the official JetPack 6.2 / 6.2.1 image and perform an APT upgrade directly on the device. [1, 2, 3, 4, 5, 6, 7]
How to Get Your System Running on JetPack 6.2.2
1. Flash the Base Image
Download the base SD Card or BSP image from the official [image]JetPack SDK page.
Use Balena Etcher to flash the image onto your SD card or NVMe SSD.
Boot your device and complete the initial Ubuntu setup. [1, 2, 3]
2. Perform the APT Upgrade to 6.2.2
Once your device is running and connected to the internet, open the terminal and run the following commands: [1]
3. Verify Your System
After the update is complete, reboot your device. You can verify your installation and core firmware versions using the command: cat /etc/nv_tegra_release
Hi, thank you for your feedback. We will synchronize the latest Jetpack device tree source code to this repository within the next two days, so please stay tuned.
Hi, thank you for your feedback. We will synchronize the latest Jetpack device tree source code to this repository within the next two days, so please stay tuned.
following up on your reply above — do you plan to adapt the A-series carrier boards to JetPack 7.2 as well? I’m running an Orin NX on an A608, and the wiki page for it still tops out at JetPack 6.2.
the Wiki hasn’t been updated for 7.2 yet, but the 7.2‑firmware is already available. You may flash the device following the steps below. The required firmware is included in this link.(608_jp72.zip) I will update the Wiki content as soon as possible.
Flashing Steps
**1. Hardware Preparation**
PC running Ubuntu 20.04 / 22.04
Type‑C data cable
A608 carrier board + Orin NX / Orin Nano module
Power adapter
**2. Download System Image and Assemble the Flashing Package**
Place the `608_jp72.tbz2` driver package and `Jetson_Linux_R39.2.0_608.sh` script in the same directory, then run the script:
```
sudo ./608_jp72.sh
```
**3. Put the Device into Recovery Mode**
The device must enter Recovery mode for flashing. Follow these steps:
Short‑circuit the RECOVERY pin and GND pin on the carrier board.
Connect the carrier board to your PC with a Type‑C data cable.
Power on the device.
Run `lsusb` on your PC. Recovery mode is confirmed if you see product IDs: (7323/7423/7523/7623).
**4. Flash the Firmware**
Flash the system to the solid‑state drive:
```
cd 606_flash/Linux_for_Tegra/
sudo ./l4t_initrd_flash.sh --erase-all jetson-orin-nano-devkit-super internal
```
The device will boot automatically once flashing completes.
**5. System Backup and Restore**
Back up the image (device in Recovery mode):
```
sudo ./tools/backup_restore/l4t_backup_restore.sh -e nvme0n1 -b jetson-orin-nano-devkit-super
```
Flash using the backed‑up image (device in Recovery mode):
```
sudo ./tools/backup_restore/l4t_backup_restore.sh -e nvme0n1 -r jetson-orin-nano-devkit-super
```
Hey, so I own the Jetson Orin NX robotics 16GB version (J401). I was able to flash it with the package downloadable from the wiki. But now I want to create a flashable image of the existing Jetson. How do I go about this? The download from the wiki does not include any tools that can be used for creating a flashable image.
So , If I may …
To clone an existing setup from an Orin NX on a J401 carrier board into a reusable, flashable image, standard JetPack / L4T tooling runs from an Ubuntu host PC over USB Recovery Mode.
The official pre-built flash zip from the Seeed wiki only includes a streamlined flashing rootfs, which is why the utility scripts appear missing. The required tools are NVIDIA’s l4t_backup_restore.sh and l4t_initrd_flash.sh.
1. Tools Required
Host Machine: Ubuntu 20.04 or 22.04 x86_64 host PC with sufficient free disk space (at least 60–100 GB depending on NVMe size).
NVIDIA Linux for Tegra (L4T) BSP Directory: The standard Linux_for_Tegra directory matching the JetPack/L4T release (found inside NVIDIA’s driver package or Seeed’s custom BSP archive).
The L4T from NVIDIA was missing DTB files. I now setup the L4T from seeed github for R39.2.0, but now I get the following error during backup
copying initrd(/Linux_for_Tegra/bootloader/l4t_initrd.img)… done.
bl is uefi
Making Boot image… done.
Not signing of boot.img
Making recovery ramdisk for recovery image…
Re-generating recovery ramdisk for recovery image…
/Linux_for_Tegra/bootloader/ramdisk_tmp /Linux_for_Tegra/bootloader /Linux_for_Tegra
75712 blocks
_BASE_KERNEL_VERSION=6.8.12-1021-tegra
command is failed
Cleaning up…
which seems to me like there is an issue with ramdisk but I have no idea what is happening…
Should I try flashing the system using the image I just built and then maybe creating the backup will work?
NO, Understand what is happening first.
The failure at _BASE_KERNEL_VERSION=6.8.12-1021-tegra command is failed is caused by an incomplete or unpopulated host rootfs/ directory in the newly setup Seeed BSP folder.
What is Happening Behind the Scenes
The Backup Script Mechanism: When running l4t_backup_restore.sh, the tool needs to build a minimal RAM-bootable recovery environment (recovery.img / initrd) on the host machine to push over USB-RCM to the Orin NX.
Generating the Recovery Ramdisk: The script unpacks bootloader/l4t_initrd.img into bootloader/ramdisk_tmp, determines the base kernel version, and then searches the host’s Linux_for_Tegra/rootfs/ directory to copy essential kernel modules (/lib/modules/<kernel-ver>/...) and binaries.
The Root Cause: When cloning or downloading the BSP repository from Seeed’s GitHub, the rootfs/ folder is empty (or only contains a README.txt placeholder). Because the script cannot find /lib/modules/6.8.12-1021-tegra or associated system utilities inside rootfs/, the copy/find command throws a non-zero exit code, terminating with command is failed.
You, should not try flashing the system using the newly built image yet—doing so will overwrite the Orin NX they are trying to back up. Instead, populate the host’s BSP tree properly:
Unpack the Sample Rootfs: Download the matching NVIDIA/Seeed Sample Root Filesystem for that release, place it in Linux_for_Tegra/rootfs/, and unpack it with full permissions:
cd Linux_for_Tegra/rootfs
sudo tar -xpf ../../Tegra_Linux_Sample-Root-Filesystem_*.tbz2
Apply NVIDIA Binaries & Overlays: Back in the main directory, apply the proprietary drivers, modules, and Seeed kernel packages to the rootfs:
cd Linux_for_Tegra
sudo ./apply_binaries.sh
Verify the Module Directory Exists: Confirm that the kernel module path matching that string actually exists on the host:
ls Linux_for_Tegra/rootfs/lib/modules/
Re-run the Backup: Run the backup command with sudo so it has the permissions required to mount/chroot/manipulate the ramdisk temp files.
Try that and report back, SHould get you straight
HTH
GL PJ
Remember to take it out of forsed recovery (remove jumper)
Hello,
I did a bit more research and apparently this is a known bug for host systems that no longer support dsa.
I took this from the NVidia Forum and it works to resolve the issue:
Suggested fix
The DSA key is never used — prepare_sshd_files() writes an sshd_config that
declares HostKey for rsa, ecdsa and ed25519 only. Simply removing the line is
correct. If you want to keep it for older hosts, make it conditional:
if ssh-keygen -t dsa -N "" -f /tmp/.probe >/dev/null 2>&1; then
rm -f /tmp/.probe /tmp/.probe.pub
ssh-keygen -t dsa -N "" -f "${_initrd_dir}/${ssh_config_dir}/ssh_host_dsa_key" >/dev/null 2>&1
fi
Separately: please pass a message to every check_error call. A bare command is failed after ~40 lines of unrelated output is very hard to trace.
Workaround
sed -i '/ssh-keygen -t dsa/s|^|#|' \
Linux_for_Tegra/tools/ota_tools/version_upgrade/ota_make_recovery_img_dtb.sh
Confirmed working — flashed successfully to NVMe afterwards.
PJ 
PJ 
