Jetson Nano 4GB Developer Kit (Official demo)

Jetson Nano is great for not only robotics/edge AI, you can use ML for science usage such as medical imaging or environmental data to speed up your workflow. From personal experience even an underclocked dual-core power save mode on the Jetson Nano will still be faster on CUDA AI/ML tasks than a Raspberry Pi 4, however your workflow may vary. If you use AI/ML that isn't optimized for CUDA, in some cases a Pi 4 raw CPU compute can edge out the Nano. I would say if you pair a Pi 4 with any AI/ML accelerator it'll cost more than a Jetson Nano and your mileage is still going to vary. Depending upon how you use a Jetson Nano, for robotics/automation you can actually run four cameras via USB and use the camera interface. Performance wise if you do opt to run a Jetson Nano using USB power, your mileage is going to vary as not all USB power adapters provide a stable voltage which means checking the specs--I reused a Canakit USB power adapter from a retired Pi 3 and never had any voltage warnings but if you plan to run a Jetson Nano hard like a Pi 4 you'll want to use the barrel power adapter for extra power stability when using multiple USB devices+GPIO. Thermal wise I've compared a fanless vs fan equipped Jetson Nano, even under sustained load the heatsink size prevents it from thermal throttling too much. This B01 version has two camera connectors which is geared for stereo imaging however you can run two cameras at a small performance loss and also fixed the networking issue which occurred on the original Jetson Nano A01/A02. From a performance per watt/dollar ratio, if you're going to dive deeper into AI/ML a Jetson NX is more ideal. With a Jetson Nano if you're pushing four cameras and LIDAR it'll require a bit of tweaking to get optimal performance and still remain at about 3.5GB of memory usage.

Very good shipping, everything arrived well, and the quality of the board is very good, as well as the efficiency to process ML algorithms

This board wipes the floor with the best of the Raspberry Pi's. It's so responsive and performant. It does not get stuttery at medium use like the Pi does. You can actually play a YouTube video on it and forget it's an embedded computer. The built-in heat sink is sufficient for all CPU usage, but if you plan to work that GPU out with heavy ML you need a fan.

After repairs from shipping damage. Not the sellers fault. It works great. I'm using it as the brain for my r2d2 project and so far it's doing everything I ask it to.

para empezar muy bueno, sobre todo en procesamiento de imagenes

I hope Nvidia keeps producing these. The price/performance point is perfect for the application I am using it in.

I expected this to be, basically, a Raspberry Pi with a GPU. A computer I could flash with any arm64/aarch64 OS and it Just Work. Boy, was I wrong. Despite being called a "Dev Kit", this is just the device itself. Nvidia just calls it a " non-production-grade" product, so Dev Kit is just a euphemism for it being a low-quality product, on purpose. The docs say you need a 5V 2A power supply, but that is not true. It will not even boot unless it has at least 4A, so make sure you buy the right one. The Jetson Linux SD card image in the Getting Started docs is half a decade old. Even the most current version of Jetson Linux available for this product is based on Ubuntu 18.04, which reaches EOL in May 2023. Nvidia is *not* releasing updates for Jetson Linux that are compatible with this product, so it will be obsolete and wholly unsupported by both Nvidia and Canonical after May 2023. So you're on your own. Even after getting Jetson Linux up and running, it's stuffed with bloatware like OS-integrated Facebook search (circa 2018) and a variety of games and a vastly outdated version of the whole Libreoffice suite. The default desktop theme and wallpaper is nausea-inducingly garish, reminiscent of the fad around lightning and dragon themed bowling shirts, circa 1998. It is possible to upgrade the OS from Ubuntu 18.04 all the way up to 21.04 if you're motivated enough. Each upgrade causes more volatility and instability, though, so YMMV and if you're planning to do this, be very cautious and make backups along the way. I spent 7 hours attending this process (yes, you have to attend it because along the way, there are random confirmations you actually have to read, because sometimes the default is to abort the process, setting you back an hour or more. Ask me how I know this). After carefully upgrading, one major version at a time (bionic to focal, focal to jammy), the upgrade from jammy to koala bricked the OS and none of my recovery attempts were fruitful. The problem with just using Jetson Linux right out of the box, without upgrading, is that you have to work within the time capsule of what was available at that time. Modern tools, AI models, etc, developed since will either not have the necessary dependencies, or taken as a whole, will not work with such an antiquated system. Combine that with this product being effectively abandoned with no future in sight for official support or continued development, and even if you're a seasoned, highly competent software engineer, you're in for a world of hurt. If you do buy this product, I hope you have a lot of patience and tolerance for disorder and chaos. You're going to need all of it.

Arrived on time. NV Jetpack setup simple and fast. Used new Samsung Pro Plus 128 sd had already. Temp stable with heatsink only (case arrives tomorrow).

J’ai eu le même problème que Thomas plus bas. La carte ne s’allumait pas. Je l’ai retourné et le vendeur devait revenir vers moi, ce qu’il n’a pas fait, le remboursement a pris 2 semaines à ce faire. De plus, il y avait une carte Kubii, un vendeur de carte jetson exactement la même mais pour moins chère. Très déçu du produit je ne recommande pas

Wi-Fiモジュール（8265NGW）とアンテナ（Econlineshop 3dBi デュアルバンド 802.11a/b/g/n/ac対応 WIFI/Wimax/Bluetoothモジュール用アンテナ MHF4 MHF4-50）を別途購入し使用していますが問題なく動いてます。耐久性は買ったばかりなのでわかりません。

Innanzitutto la scatola non è nvidia e questo già mi ha fatto pensare… poi nella descrizione viene specificatamente inserito il codice che si riferisce all’utilizzo della sd, mentre questo modello usa un’emmc da 16 gb, i quali non sono nemmeno sufficienti per installare il sistema operativo ed i pacchetti nvidia… spiacente per il ritardo della recensione ma non ci è voluto poco per capire che il problema non fosse al boot o di tutto il resto ma proprio della scheda arrivata sbagliata

Carrier board logic level converter is defective on all boards; the 3.3V output is 1.8V and randomly switches. Pinmux flashing doesn't fix and if you have any noise in your system the GPIO will flicker on and off rapidly in response to the noise. Inputs require pull down resistors of 1.5kOhm to drain 3mA to ground in order to function and not float high at 1.2V (THESE ARE INPUTS). Sometimes the outputs will randomly change from 1.8V to 3.3V. This is occuring on multiple jetson nano boards and not isolated to one device. The carrier has defective hardware.

Great product