Nvidia Jetson Nano – the 99$ Tegra Computer running Linux
Following in the wave of the raspberry pie and other single board, Nvidia released its own board, called the Jetson Nano, in early 2019. They were kind enough to send me one, and I’d like to see what can be done with it ! Let’s take a look !
Support the channel on Patreon: https://www.patreon.com/thelinuxexperiment
Follow me on Twitter : http://twitter.com/thelinuxEXP
What I use to make my videos:
Microphone: https://amzn.to/2PsNWXl
GPU: https://amzn.to/2LHZ5o5
Motherboard: https://amzn.to/2KZt63t
CPU: https://amzn.to/2IFjKrw
What’s the nano ?
The jetson nano differentiates itself from other single board computers, in that it’s means primarily for AI and machine learning, and as such, ships with a good GPU onboard, using that power to crunch data as fast as possible. Obviously, you can use it for other purposes, with one major caveat we’ll talk about later.
Nvidia provides an SDK to get started and start building your own AI-based applications.
It’s also not a true single board device, since there is the actual nano itself, this small module here, and the PCB which is just intended to provide IO to the Nano.
This device is specifically important, since it’s a cheap gateway into the world of AI and Machine learning, which can require pretty expensive hardware to get started.
What’s in the box
I can’t say exactly what will ship if you order one, but here is what I got in mine !
First is the nano itself, a pretty small, but high naked board with the necessary inputs, like a micro SD card slot for storage, 4 usb 3.0 ports, hdmi, display port, and a 40 pin expansion header. It also has a camera connector, and a gigabit ethernet port.
You can power the nano through micro USB, or the dedicated DC jack.
The Nano does not come with a wifi chip, though, but you can add an M.2 dedicated module, which should make the nano a lot more appealing for uses outside of the machine learning space.
I also got this fan, that will be needed to cool down the GPU when the Nano is crunching data.
Finally, there was the SD card itself, which contains the OS, which is, obviously, Linux.
My nano came with a power cable, but this, the SD card, and the fan are not part of the package Nvidia sells, so you’ll need to add at least a micro usb power supply for turning this thing one. The heatsink provided might be enough for a start, but you might want to invest in a fan if you plan to really use this thing.
Obviously, you’ll also need a keyboard and a mouse, and a display to set it up.
3 – The specs
I’m not big into spec sheets, but the nano, for its 99$ price, packs a lot of punch. It uses a quad core ARM A57 CPU, running at 1,43ghz, 4GB of DDR4 RAM, and its GPU is based on the nvidia Maxwell architecture, with 128 cores, which total 472 Gigaflops of computing power. For those who lack a frame of reference, this puts it at a third of the original xbox one’s computing power, which is not bad for such a small form factor and considering its price.
It’s able to decode up to 4K60fps video, and encode up to 4K30fps, which could also make it a pretty useful dedicated game streaming machine.
4 – What does it run ?
By default, the OS loaded on the SD card is a variant of Ubuntu 18.04LTS, with a kernel compiled for ARM processors, and a bunch of libraries useful for the nano’s intended purpose.
It uses XX Gb of storage, and ships with some programs out of the box: LIST OF PROGRAMS.
The distro is pretty lightweight, using only XX out of the 4GB of RAM shipping with it.
It’s still an ARM version of Linux, so you won’t get all the software you might already be used to, but it does run a GUI, so you could theoretically use it as a standard computer. This brings us to the next point…
What does it do?
Nvidia markets this device heavily as a Machine Learning and AI developer kit. For the price, it’s a pretty good device, that allows anyone to get started and learn to develop AI and machine learning projects. It ships with a bunch of test projects and sets of data you can use to train yourself to code ML and AI related programs, or even create an AI robot.
This is not all it could do, though. You could use it to run a nextcloud instance, a web server, or even as a streaming device and video encoding machine. It could be used as a game emulation machine, a NAS, or as a Kodi or plex server, although these other uses could also be served by a less expensive raspberry pie.
As with all single board devices, there isn’t much this thing won’t do, and we are going to take a look at how to use it for many different use cases, as well as the machine learning tools Nvidia provides, so stay tuned for more !
by The Linux Experiment
linux web server
