JetPack 4.6.3/L4T 32.7.3 released

JetPack 4.6.3 with L4T 32.7.3 is released. JetPack 4.6.3 is the same as JetPack 4.6.2 but includes Jetson Linux 32.7.3 which brings in security fixes. It also adds support for Jetson TX2 modules with new Hynix memory.

This release supports Jetson AGX Xavier series, Jetson Xavier NX series, Jetson TX2 series, Jetson TX1, and Jetson Nano.

there is any guidance for installing/updating Jetson TX1 ?!
I have another version of JetPack on my Jetson Board but now I want to install or update it again.

Each release has its own documents, but you might find this of use:
https://forums.developer.nvidia.com/t/jetson-nano-2gb-wont-boot-beyond-nvidia-splash-screen/220018/5

That URL has the URLs for flash software download (you’d use an Ubuntu 18.04 host PC).

Is there is any way to flash the software of jetson without using host PC?
I can’t use the host PC for doing that…
can I use an SD Card instead of using the host pc?
.
tomorrow I just run command:
sudo apt update
sudo apt upgrade
and then reboot my Jetson, after rebooting nothing has been installed on jetson, and didn’t run ubuntu and GUI…
I just can’t be using the terminal…
I don’t have any idea what’s going on and what can I do???
when I power on Jetson I get these Errors:

The software which is flashed can be roughly divided into boot content and rootfs content. If you have an SD card model (not an eMMC model with an extra SD card on the carrier board, but a true dev kit), then the boot content is on the QSPI memory of the module. The rootfs (o/s) would be on the SD card.

Jetsons do not have a BIOS, and so they cannot self-flash. The content which brings up rails and clocks prior to the bootloader could be considered roughly equivalent to the BIOS of a PC. When you flash a PC you don’t simultaneously flash the BIOS. That BIOS is why a PC can self-install (it is its own “equivalent” to a mini operating system). Flash in recovery mode is mandatory for the QSPI content, and in recovery mode the Jetson is equivalent to a custom USB device (a recovery mode Jetson is not a mass storage device). The software which understands that custom USB device is the “driver package”, and this runs only on a Linux PC, so there is no way around it.

The content in the QSPI must be compatible with the content of the SD card. It is true that “often” the QSPI content is compatible with a couple of releases on the SD card, but quite often one must flash the QSPI via a host PC in order for the SD card to correctly function. Command line flash works with a number of releases, and does not necessarily have the same requirements as the JetPack/SDK Manager GUI frontend (which uses the driver package as a backend; manual command line flash directly uses the driver package). For a TX1 you would need an Ubuntu 18.04 host PC (you could also get away with Ubuntu 16.04 on older releases). This is a GUI requirement, not a driver package requirement.

The driver package has Linux binary executables in it. Windows and other o/s cannot run this. Additionally, it depends on loopback, and any emulation without this cannot succeed. In the case of VMs (which are not officially supported) they might work; it just depends on whether you’ve correctly passed through USB (USB disconnects and reconnects during flash, and VMs tend to fail the reconnect, which is out of the control of the flash software).

Sometimes you will get the errors you see in the screenshot due to mismatch of SD card content and boot content. Sometimes you will see this due to a device tree mismatch, e.g., using a dev kit device tree on a custom carrier board.

Note that eMMC models also have separate boot and o/s content, but you’ll see that in partitions of the eMMC.

I think you probably have to flash the QSPI with the same release which was used to produce the SD card. Even if that is not the cause, then this would still be a mandatory first step.

Thanks a lot for your descriptions, it’s very help full.
do you have any guide or introduction of Ways to install jetpack and Nvidia drivers for Jetson TX1? step by step?
SDK manager?
using host PC for flashing?
using an SD card for flashing?

The frontend GUI is JetPack/SDK Manager, and what actually gets flashed is L4T (“Linux for Tegra” is Ubuntu plus NVIDIA drivers). The L4T release and JetPack/SDKM releases are tied together. Specific instructions come with the release. To see L4T releases and their compatibility, plus JetPack releases (the L4T page will take you to the JetPack content, and the JetPack content will take you to the L4T content, so just find one that matches your current setup), see:

If you install the .deb of sdkmanager, then run it as a regular user (don’t use sudo), much is shown as you go. Sometimes it isn’t obvious that you can uncheck many things, e.g., you could choose to just flash the Jetson and not install anything else to the host PC. You can also exclude installing optional packages, and then at a later time, run sdkmanager again, but uncheck flash, but leave optional package install enabled.

If you want the sdkmanager to start with availability of older releases, but still install perhaps the latest JetPack 4.x release (5.x won’t work without Xavier or newer), then you can start it like this:
sdkmanager --archivedversions

Be sure to have plenty of disk space. Also, you only enter your local PC’s admin password when asked by JetPack. You’ll need to have your email browser on that PC so it can authenticate.

Instructions on the specific release will show what you need for SD cards and QSPI flash. Should the Jetson be properly connected and in recovery mode, then JetPack will mostly self-guide the flash. Should it come time to install optional extra packages, just boot the Jetson normally and log in. Those optional packages are installed over the network to a fully booted Jetson, and not in recovery mode.

1 Like

@linuxdev I have already installed JetPack 4.6.3 using Host (VMware Machine) with SDK manager and also flashed the Jetson TX1. Thanks for your help 🌸👍
and I have another question, how can I convert Tensorflow Object Detection Model (SSD-Mobilenet) to tensorrt for inferencing on Jetson?

Beware that a VM often has USB issues since a Jetson disconnects and reconnects during a flash, but if it works, it should be ok.

I can’t answer the AI questions. That’s probably something for @dusty_nv.

yes, you are right, jetson board disconnects during flashing with the host, but this issue does not cause a problem and the board flashed successfully.
Thanks for your help 🌸🌸👍