Do I have to flash Jetson TX1 by using JetPack?

I just had a new Jetson TX1. There is Ubuntu 14.04 installed by default. Do I have to flash Jetson TX1 by using JetPack to use API and toolkits? Or they was installed by default.

Thanks in advance.

The TX1 comes with the GPL-licensed pure Ubuntu 14.04LTS installed. In the home directory of the default “ubuntu” user is the “NVIDIA-INSTALLER” subdirectory. If you run the “installer.sh” program as root (meaning “sudo ./installer.sh” while in that directory), the nVidia-specific files will be installed. Those files expand the Jetson to full hardware acceleration, which is a prerequisite for most of the interesting software, e.g., CUDA. This does not install actual CUDA. You would reboot after running that file, and perhaps install additional software (e.g., CUDA).

If you were to flash, you could use JetPack…this is bundled software with things you might want on both the host and the Jetson. Packages of JetPack could be installed manually one at a time in the proper order. If you were to install manually, the flash software is the driver package, and at minimum would also require the sample rootfs package. The sample rootfs is identical to the Jetson as it first arrives, prior to running the NVIDIA-INSTALLER software. During manual setup of flash there is an application “apply_binariers.sh”, which would unpack the nVidia-specific files into the sample rootfs…which during flash would propagate to the Jetson. The pristine Jetson+installer.sh is exactly equivalent to manual install of sample rootfs+apply_binaries.sh.

Release R23.2 should be coming out in a few days, you may want to wait for that to flash.

L4T 23.2 has been posted here: https://developer.nvidia.com/embedded/linux-tegra
The Jetpack bundler for it will be uploaded shortly, it automatically installs CUDA toolkit, OpenCV4Tegra, VisionWorks, ect.

If I was using stand alone TX1 modules for use with my own motherboard, what mechanism would I use to go about
flashing my own customized BSP onto the module (e.g. at the factory)?
JTAG is traditional for many SOCs and uCs, but it would be very slow.
What are my alternatives and is there some documentation discussing this? For instance could I flash the EMMC
via USB port?

As on the reference devkit, you would flash the module by resetting it into recovery mode and attaching a host machine to USB port #0, then running L4T flash script or Jetpack. If your custom carrier/motherboard is not designed to include USB port, then recommend flashing module using reference devkit on your factory production line.

Hi, I just flashed my jetson TX1 with Jetpack 2.1 and after its done flashing OS its not able to determine my TX1’s IP address. I connected my jetson to a monitor and I only have a command line interface and when I try to login it dosent accept ubuntu for login id.

On command line, does anything not show as ok for:

sha1sum -c /etc/nv_tegra_release

@geoffreywall how did you flash the image on the module finally?