Unpacking the nVIDIA AGX

So I ordered the https://developer.nvidia.com/embedded/jetson-agx-xavier-developer-kit. It arrived today in mail and I unpacked the beast. I am very excited to get me some Deep Learning with CUDNN library.

Note: Please have a dedicated HDMI screen connected to the Xavier kit while installing, without this you will have a lot of headache installing.

Installation Hiccups

In the past, I have spent a lot of time playing with boards and Linux Kernel bringups, installing full Linux distro on things I shouldn’t be without which this could be a bit of a problem mainly because AGX was supposed to come with Linux and all the related packages pre installed, but mine simply refused to boot up!

NVidia has a massive website dedicated for discussions and guidance on installation. Since I bought the fancy Jetson Xavier AGX kit (which is the pricier high end kit) I assumed it came installed with the latest OS too. Turns out, all the instructions websites asked me to stop whatever I was doing and go straight through the OS installation steps, sigh!

In general, I despise installation through GUI. Sure, I have to install everything through the sdk manager software which has an added “joy” since it invents new excuses to keep balking. I quickly figured it’s just a shell script with a GTK widget window in the front. Then I was able to find my way into getting it to start installing.

Android Filesystem

The user partition on the eMMC is nothing but Android-esque bootloader, kernel, system partition.

Thermal!

Yes, the Xavier AGX is a massive power packed box but the thermal dissipation is substantial, so much that I could feel the warmth physically while sitting a good 3 feet away from it.

Install Ubuntu

Finally, after installing all the components through NVidia’s SDKManager, I installed Ubuntu on the board.

Coming up … Task Based CUDA work