NVIDIA Jetson Nano Developer Kit

A few notes on the Jetson Nano from the start: 1. - The Jetson Nano, despite it's likeness to other Single Board Computers, it is categorically different than other SBCs with an ARM SoC. Indeed, the Jetson Nano is a System on Module, and is specifically built with Intelligent Systems design, Machine Learning, Robotics, etc., as its primary purpose. 2. - The Jetson Nano, while quite capable, is not meant to be a set top box by any means - if that is what you are looking for, the Nvidia Shield TV is a rather well developed platform and would be significantly more satisfying for the home theatre setting and at a fairly similar price (the Shield comes w a Power Supply, Internal Flash Storage, WiFi and Bluetooth, a custom build of Android TV, etc.). I pre-ordered the Jetson Nano a few days after Nvidia announced its imminent release - after approx. 3wks or so, I finally received it. I had downloaded the Jetpack image file and flashed it to an SD card in anticipation of its arrival - so, setup was fast and simple. The Nano currently has Ubuntu as the primary OS, & while I am not a fan of Ubuntu, it is the cleanest OS I've encountered on an SBC, next to Raspbian and the Raspberry Pi. Compared to the Rock64, the Tritium H5, the Odroid XU4 etc., getting the Nano up and running whilst being fairly stable, the Nano is probably the easiest setup I've experienced in a while. Conversely, given the board's purpose, a ML learning platform, it has been a challenge for me for different reasons - but nothing I didn't expect. As for the board, it does not come with a power supply and it can accept power via micro-USB, through the carrier board pinouts, or through a barrel jack. It is meant to run at 10W in default mode, but is capable of a 5W mode. To operate the board at 10W, do not power the board via micro-USB. If you do and add peripherals, the board will crash rather easily. I used the same 5V/4A power supply I ordered for my Odroid XU4 and it works perfect (you will need a jumper - pictured - to select how you will power the board). The Nano requires a mSD card like most SBCs - a UHS-I, U3, Class 10 card is needed to get up and going properly; however, with 4 USB3 ports, I transferred my install to a spare SSD and it easily outperforms the mSD card. Also required - a WiFi/Bluetooth dongle or a PCIe Key A/E card, which can be installed under the module. Without, you will be forced to use the onboard Gigabit Ethernet connection. The pictures seem to make the board look somewhat large - and while it is bigger than the RPi standard, it is still fairly small. It's approx. the same length of an 2.5" SSD and slightly larger compared to the width of an SSD. The Module does have a large heat sink - again, it appears to be much larger than it actually is - the heatsink mounts a 40mm x 40mm fan for perspective. So, I know Nvidia has lost popularity over the last few years due to their GPUs; however, I have to admit, the Jetson Nano is a really great deal. Even if used as a standard SBC - the Nano is a great deal (the Shield is even better a deal for that though imo) compared to many other boards that cost the same or more. The benefit with the Nano is the access to the Jetson Package and a platform to learn and test the Cuda software. It will even have demos to see the Jetson's capabilities that come with the Ubuntu install. Again, these are just my first impressions of the board - compared to several other SBCs I own, I can already say the Nano handily bests all of them. The RPi3B+ with its community and price point is also a good deal, but it is an entirely different type of learning platform. I would say if the Nano has sparked your interest and you're not expecting an even better Shield TV...and if you are adept with Linux, then the Nano is a great deal. I would definitely recommend giving it a go.

I have other ARM PCs from Odroid XU4 and Raspberry Pi2, and neither compare to the Jetson Nano. The Pi's have excellent support, but are vastly underpowered. The Odroids have power, but HardKernel can never make a stable Linux drivers to go with it. Now here comes the Jetson Nano from NVidia, the top GPU manufacturer. The Nano comes with a fully supported LTS Ubuntu distribution fully customized and supported for the Jetson Nano. There are no incompatibilities nor issues. It also comes with a full developers kit for Artificial Intelligence development. Why? Well this sucker has an NVidiia GPU onboard. It's a Tegra on steroids. The community already have projects for facial recognition, self-driving robots, etc. It's quite the little board. It boots off the SD Card, but also has a mini-PCI-E slot under the module. The distribution of Ubuntu is "minimized" to have a small storage footprint, but you can "unminimize" it to a full blown Ubuntu desktop distribution to restore any Linux utilities you want. So, if you are wondering where all of the man pages went, well you have to unminimize first. sudo /usr/local/sbin/unminimize Voila, it's a normal Ubuntu distribution with all of the Jetson Nano extras. The GPU is exploited in the GStreamer app for video, if you want to watch 4K video. It is compatible with the second version of the Raspberry Pi camera, but I recommend getting a better one anyway, but it will work with that one, if that is what you have. The expansion bus is fully Pi compatible, but cards plug to side, and not on top of the board (due to the large heat sink being in the way). It runs fine off of the USB power port, but if you are going to be doing some heavy computing, then I recommend powering it from the power plug with a 5V @ 5 amps supply.

Awesome product, really sped up my neural network processing speed

Very very nice bare bones development board. I'm using it for telemetry processing and graphic visual rendering of such live data. Software has gotten better than when I was using the Jetson TX1 couple years ago. Very easy and quick to get going. nVidia did us right with this one. For anyone that is curious I used a 32 GB microSD card and during install it auto-expanded the root partition to utilize the entire card. I also attached a WD 314 GB USB HDD, formatted ext4, for project files compiling and large media storage. And using the barrel power connector instead of microUSB (via power bypass jumper).

First of all, the hardware is awesome. Especially for under $100. And once you've got a trained model running, it's fast, fun, and inspiring to work with. But for any newcomers to the NVIDIA AI ecosystem, there's a lot of proprietary lingo to learn and it can take some time to understand what NVIDIAs tools are actually meant to do (the NGC cloud for example). With that said, it's more of a documentation issue at this point because once you've set up your toolchain it's very straightforward.

I simply like this a lot. It provides a nice entry level to dev for embedded parallel processing boards, think a little bit ahead about how you might expand (scale out) to several (many) of these boards working together to crunch a problem, the dev kit is just where you start with this. Just one slice, having fun, prototype like crazy and try to interface to other hardware, that's going to open up your concepts of where you can go.

The order is done very well with the shipping. The product is what we wanted and meets our expectations. We are talking part in a robocon competition and this purchase will help us in computer vision

Schöner kleiner PC mit gut angepassten Betriebssystem auf Ubuntu Basis. Ermöglicht einen einfachen Einstieg für einfache KI Anwendungen. Für den Preis gute Leistung und geringer Energieverbrauch. Da lassen sich schon einiger Prototypen draus basteln. Für manche Anwendungen ist der RAM recht klein, dann lieber für eine der größeren Jetsons entscheiden. Es gibt viel Zubehör.

opencvをホビーでやる分にはラズパイ4より速くていいです。CUDAも使えるし。その他含めた汎用性はラズパイの方がありそうですが。 私が買ったのは旧リビジョンですが、最近新リビジョンが売られています。商品写真は新旧混在しているのでどちらが送られくるか確信もてません。情報が追加されるのを待った方がいいかも。

Awsm product .... at first got a defective one ( defect in its barrel jack) but then got the right one after a replacement ..... thanks to Amazon

Likes 1. 128 Nvidia graphics cuda cores 2. Easy to download operating system and install from Nvidia website 3. Very fast even from a memory card 4. Lots of documentation 5. Machine learning examples 6. Gigabit ethernet is faster than gigabit ethernet of PC 7. Temperatures never cross 45 even under moderate load, thanks to bundled heatsink 8. Excellent full fledged User Interface 9. Kernel is like real-time and everything feels snappy 10. 64 bit arm, wow!! 11. Excellent power management, runs in low power mode if you supply less power and runs like pc if you supply additional power which is well documented 12. Man it supports 4k monitor, kudos nvidia Dislikes 1. No case or proper stand, no screws or latch, I have to build my own case by drilling holes in a plastic hard disk case to convert into a stand, used old ssd screws to tighten the motherboard 2. You are stuck with Ubuntu provided by Nvidia,not as open as raspberry pi or intel nuc which allows freedom to install any operating system of your choice 3. No wifi or Bluetooth 4. No jumper provided which is required to use power other than usb 5. Installing tensorflow takes forever as it builds all the packages on jetson, ideally they should be downloadable binaries 6. Missing Hardware Acceleration Since its arm64 chrome or firefox doesn’t support hardware decoding of videos, you are stuck with the arm processors eventhough there are 128 graphics cores 8. Hardware acceleration for video decoding is a closed box, very little documentation to achieve hardware acceleration for videos using kodi/vlc/mplayer