While Raspberry Pi boards are great for doing all kinds of tasks and they're capable of doing object recognition, they can be a little slow when it comes to real-time image recognition. In 2019, Nvidia came out with an A.I.-focused Pi competitor in the $99 Jetson Nano. Sure, the 4GB Nano had four times as much RAM than the top-level Pi at the time, but it was more than twice as expensive and didn't come with Wi-Fi.
Fast forward to 2020 and Nvidia is back with a 2GB version of the Jetson Nano that sells for a more reasonable $59 and, for consumers in some markets (including America), comes with a compatible USB Wi-Fi dongle in the box. Due out later this month, the new Nvidia Jetson Nano 2GB is designed to make A.I. more accessible to hobbyists, kids and aspiring developers.
To help more hobbyists make use of its platform, Nvidia is also introducing free online training and certification programs for A.I. Since the nearly-identical 4GB Jetson Nano has been on the market for a year and a half, there's also an existing community of developers who've shared tutorials and open-source projects. You can even buy Jetson robot kits to build.
Like the 4GB Jetson Nano, the new model is powered by a 64-bit, quad core ARM A57 CPU running at 1.43 GHz, along with a 128-core Nvidia Maxwell GPU. According to Nvidia's numbers, the Jetson Nano 2GB scores anywhere from 8 to 73 times higher than a Raspberry Pi 4 on A.I. benchmarks such as ResNet, OpenPose, Inception V4 and Tiny YOLO V3. While we haven't gotten to run these tests ourselves, it's easy to believe as the 4GB Nano was significantly faster at A.I. than a Pi.
Nvidia Jetson Nano (2GB) Specs
Swipe to scroll horizontally
Header Cell - Column 0
Jetson Nano (2GB)
MSRP
$59
CPU
64-bit Quad-core ARM A57 (1.43 GHz)
GPU
128-Core Nvidia Maxwell
RAM
4GB DDR4
Connectivity
Ethernet 10/100/1000, Wi-Fi dongle (some regions)
Ports
2x USB 2.0, 1x USB 3.0,HDMI, CSI Camera Connector
Pins
40-Pin GPIO
Jetson Nano 2GB vs Jetson Nano 4GB
In an improvement from the 4GB model, the 2GB Nano gets power over USB Type-C where the older unit had a proprietary barrel connector. True the older model comes with a 5V, 4A power supply (it could also do 5W over a micro USB connection) and for the new one, you'll have to bring your own USB-C charger of at least 5V, 3A, though many of us already have drawers full of them.
In a less-welcome change, the Nano 2GB has fewer USB ports, with only one USB 3.0 and two USB 2.0 ports versus 4 USB 3.0 ports on the 4GB model. It also has just HDMI out while the more-expensive model also has a DisplayPort connector. However, bundling the Wi-Fi dongle is a huge advantage for anyone who wants to use the board and isn't near a router with Ethernet out. The 4GB model doesn't come with a Wi-Fi dongle and, in our testing, we found many models that didn't work with it.
(Image credit: Nvidia)
Jetson Nano vs Raspberry Pi
Like any recent Raspberry Pi, the Jetson Nano 2GB has a 40-pin GPIO connector and a CSI camera port. The camera port is compatible with Raspberry Pi camera modules, which is good news for anyone wanting to build a robot or iOT device that uses image recognition. From a GPIO pin-out perspective, Raspberry Pi HATs (including the best Raspberry Pi HATs) could and should work, but the software to take advantage of individual HATs may or may not exist for Jetson.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Where the Raspberry Pi's official operating system is Raspberry Pi OS, a port of Debian, Jetson Nano's is eLinux, a version of Ubuntu. So, while you could probably get a lot of the same software for Nvidia's board, you can't just download every software package that’s meant for the Pi and use them without modification.
Bottom Line
It’s doubtful that the Jetson Nano 2GB will take mindshare away from Raspberry Pi, but it does have a chance to gain critical mass with makers who want to build more A.I.-intensive projects. For example, Maker Ahad Cove recently designed a self-driving trashcan that knows when the garbage truck is coming down his street and drives itself to the curb just in time for pickup.
Cove uses a Raspberry Pi to power the trash can itself, but the computer which looks out his window and sees if each passing vehicle is the garbage truck is an Nvidia Jetson Xavier NX developer kit, a $399 model that’s quite a bit faster than the Jetson Nano, but uses similar technology.
Cove said that he has used Jetson Nano in the past for object recognition projects and, when we asked him why he doesn’t use Raspberry Pi, he noted that Pi is just not quick enough to detect the truck in time to send the trashcan out to the corner. It’s this kind of local object recognition where the Jetson Nano 2GB could come in very handy for hobbyists who don’t have hundreds of dollars to spend.
Avram Piltch is Managing Editor: Special Projects. When he's not playing with the latest gadgets at work or putting on VR helmets at trade shows, you'll find him rooting his phone, taking apart his PC, or coding plugins. With his technical knowledge and passion for testing, Avram developed many real-world benchmarks, including our laptop battery test.
It's a good new for development, but how much will cost the compute module? .
The problem with the developer pack, is that you can't use it to create and sell a product, to do it you must buy the compute module and design your own mainboard to host the compute module, then it will cost more than us$500.
I hope Nvidia will sell a complete solution for hardware vendors like me, with an affordable price, and not just a development kit .
For now I will continue working with a raspberry pi because it cost 1/5 of the price of a solution based on the jetsony nano compute module.
