we came into posession of some google jamboards which will be EOL next year. Since they rely on specific network services from google they’ll be useless by then. And since they’re based on a Jetson XT1 we thought we could just install another os.
However, the latter turned out to be tricky since we have no experience with such devices. Looking at the sdk downloads we found that SD card images are not available for the XT1 but apparently require the installation of the sdk on ubuntu from which we then could prepare an SD card.
In short, any help with installing a different os onto a Jetson XT1 would be very welcome!
I’m not sure what the google jamboards is, supposed that’s not the devkit, and customized product by 3rd party.
You should discuss with them for how to do the SW update.
There is no support on a TX1 other than that provided for the specific release of Ubuntu. Jetsons have a custom boot chain, and no BIOS. This means that the equivalent of a BIOS is performed purely in software, and that this is not “abstract”…only the environment flashed by the specific board support package will work. Even if the Jetson TX1 module is used, a custom carrier board would require a custom device tree, and that is for the supported Ubuntu.
TX1 is very old. It isn’t until Orin that there is any support for other operating systems, and that is with L4T R36.x (the first R36.x stable release for Orin is quite recent, and still requires some adaptation to get other content installed). TX1s had their last release at L4T R32.x. Note that L4T is just Ubuntu plus NVIDIA drivers. See: https://developer.nvidia.com/linux-tegra
However, the currently installed os works fine with the original board as well as that Orbitty carrier I’m using at the moment. So, once installed, the board doesn’t seem to matter anymore and hence I was hoping I could install the os again. Even if it’s just the old, officially supported ubuntu. The goal is just to get a fresh start, no matter with which os.
The URL I gave in my previous post provides the flash software for NVIDIA development kits, and documents on those specific L4T releases (you’d want the most recent L4T release that goes with your model).
Do be careful though with third party carrier boards. If you use the wrong flash software, then you will no longer be able to boot. Jetsons have options to which functions appear on which pins of the module, and third parties might design a carrier board that (A) is an exact electrical layout match to the dev kit, or (B) has layout modifications. If the layout is an exact match (electrically), then the manufacturer will likely just state that you use NVIDIA’s software; however, if the second case is correct, then either the manufacturer will provide a patch to the NVIDIA software, or else the manufacturer will provide its own rebranded release of that software with modifications. Any layout differences between the wrong firmware used and what is correct will result in parts of the system not working (including perhaps not booting).
Okay so I extract Tegra_Linux_Sample-Root-Filesystem_R32.7.4_aarch64.tbz2 onto an SD card, put that card into the carrier and then the Jetson should boot from that?
No. There is a lot more to creating a filesystem than just unpacking those files. JetPack/SDKM can download and install without actually running and flashing in some cases. However, if you want to do this manually, consider these steps:
Download the sample root filesystem (which you mentioned). Also download the “driver package”.
Unpack the driver package as a regular user (do not use sudo). This will produce a subdirectory, “Linux_for_Tegra/”. Almost all commands originate from the Linux_for_Tegra/ subdirectory.
Now go to “Linux_for_Tegra/rootfs/”. This location is empty other than perhaps a README type file. From there use sudo to unpack the sample rootfs into this location.
Now cd back to “Linux_for_Tegra/”, and run the command “sudo ./apply_binaries.sh”.
At this point you can use command line flash and do not need to recreate any of those above steps again.
There will be documentation with that release. This will tell you how to create an SD card for this. Beware that this will consume a lot of disk space every time you create an image. You could of course go to that URL and use a pre-created SD card image of that release.
The other half of this is that there are two types of Jetson TX1 Nano modules, and we don’t know if you are using one of those, nor if you are using a full TX1. It isn’t possible to be certain about many things because this Jamboard is an unknown carrier board. One thing you can do though is to look closely at the module itself.
Is the module in the form of a DIMM memory module?
Or is the module a more rectangular unit with heat sink and a non-DIMM style larger connector under it?
If this module is a DIMM style format, does it have an SD card directly on the module? If so, then this should have been with a dev kit and is not really supposed to be used in third party products. If the module is in DIMM format, and it does not have an SD card slot anywhere on it, then it is a commercial eMMC module. The steps for each differ (and so do carrier board steps, and we know nothing of the carrier board).
If you can identify the module, then perhaps we could offer suggests such as an SD card model having the QSPI flashed. If it is an eMMC model, then you won’t use an SD card. If there is boot code for some other device other than a module-integrated SD card and other than eMMC, then odds are much higher that any flash we provide makes the unit fail. Firmware matters during boot, and the primary carrier board difference is firmware. You might have some success if you find the correct module type, but it is risky.