How to install Jetpack 4.2 installation from SSD in NVIDIA JETSON TX2 dev. kit

presently I am working with NVIDIA jetson TX2 dev. kit with Jetpack 3.3 for my video processing research. I am using it as an edge device. Also using one SSD of 128 GB of WD. Now, I want to change the software to Jetpack 4.2 with facility of booting the OS from the SSD so that I can Ihave more memory for storing and executing programs. This is required as I need more memory for my deep learning based applications.
Can anyone tell me how to start developing in Jetson TX2 using SSD. I have also watched similar video regarding “NVIDIA Jetson SDK Manager - JetPack 4.2”. Only missing part for me is how to make setup for the development from the SSD. They have already made a similar video for Jetpavk 3.3, i.e. “Develop on SSD – NVIDIA Jetson TX1 and Jetson TX2”.

Otherwise, can suggest me how to change the root files in the SSD to make the changes happen so that I can develop on SSD? One more doubt is, are all the jetson TX2 boards are compatible with Jetpack 4.2? Kindly revert back me with necessary suggestions.

Apart from this I need to know that can I install cuda tool kit 10 in NVIDIA Jetson Tx2? in other-words, Is jetson TX2 is compatible with CUDA 10 (turing architecture).

I don’t have an SSD to test with, but probably the easiest path to choose is to flash the TX2 normally to SDKM4.2. Then look at this thread with a similar question:

I don’t know if U-Boot is capable of booting from the SSD directly. If it is PCIe, then probably not. If USB SATA, then probably it can. Someone else will need to answer that. The previous thread I mentioned does not have that question due to knowing that U-Boot can boot from an SD card.

To boot from a SATA-SSD connected via one of the USB 3.0 ports you have to

  1. Recompile the kernel so /lib/firmware/tegra21x_xusb_firmware is part of its Image
  2. Rsync the microSD card to the partitioned SSD
  3. Update the boot config /boot/extlinux/extlinux.conf reference the SSD for boot contains a nice and detailed explanation contains an automation of this approach - assumes Ansible knowledge

Thanks for sharing!