Gift for Mother Day

Ok, not exactly … ehm…
But arrived saturday just in time …

Fez Hat still working under Ubuntu arm64 netcore 3.0 with GPIO, Waw!
Next test I2C …


Is this board any good? I am considering getting from from Seeed Studio as they still have stock left.

How easy is it to programme? What you using to develop with it?

Dave, the board is good for the price, although not exceptional (TX1/TX2 are lot better). Nvidia wants to get some more “advanced maker” market for sure. Very poor documentation, apart from the AI demos. Not even know how the gpio pin are named/numbered, except a pinout sheet made by a good will guy not from nvidia… It’s not even clear how to power supply the board, so I’m using 4A 5V barrel plug, but it seems you can use 2.5A (???) from micro-USB.
It’s really a lot faster compared to PI3+ with really 10-15x fast framebuffer (oh… well that’s nvidia core business). OS is Ubuntu 18.04 AARCH64 re-imaged by Nvidia with tools named JetPack.
The good news is that is Netcore supported starting from Netcore 2.1 arm64. ASP net core is supported from 3.0 (now preview5) and later.
There’s no Wifi/BLE on board by default, but you can buy Intel M.2-E, a $25 board interface to implement it, plug&play.
I must say that it’s unbelievable powerful coming to Tensorflow and Inference AI engine… It can process 11-15fps shape recognition from a webcam…

We are asked to R&D a system for automation on a meat preparation line, collecting lot of infos during process (quite a big engineering project by an industrial automation firm). Now Nvidia Jetson can help for visual inspection part of the project.


Just to show some result of weekend tests…
I found that:

  1. SPI bus not mapped to connector due to Nvidia device tree: it is not enabling SPI pins on 40pin connector. I could modify DTS but it needs to be flashed on internal nvram (it sucks becouse nv_flash program need Ubuntu 18.04 and I have 19.04… I need to install a new VM).
  2. GPIO input need stronger pull-up in the order of 5Kohm or less (10k is fine with RPi, but left floating the nano), and it seems that internal pull-up setting is ignored by input direction. This could be a compatibility problem for Pi Hat
  3. I2C is working fine … I could write (port) some driver to .NET core arm64 to use FEZ_Hat chips.
  4. I could run “as is” lot of my .NET core software (also ASP.NET core) without any problem.

That said, still remain an interesting platform.