Seeed Studio A205 Carrier Board for Jetson Nano, Jetson Xavier NX with Compact Size and Rich Ports. (6 CSI Camera, 2 HDMI, 5 SATA, M.2 Key E Supported etc.)

When I first purchased this, I could not get it to power up. My eyes are bad and I thought the jumper came in the right position for a 5v 4a power supply. Please note to pull it out and cover the other pin. Do not attempt to run this with a lower amperage power supply, it will get about half the performance or worse.Also, make sure you get a fan. This thing gets warm and the GPU will slow down if it gets hot. The fans are a major pain to screw to the heatsink but worth it as it gains about 15-25% performance.Be aware that you must use Nvidia's tailored Ubuntu OS. I tried to do a custom install but it would not work.I am using this to do AI predictions. I started with 3 small networks. It is taking about 10ms per network. You would not use this module for training, but for model usage it is amazing. Don't let people steer you away from this if you are doing small networks.If you are trying to capture full-motion video and predict at 60 times a second, then get a Xavier. If you are doing more intermittent detection, say every 50ms, then this will work.For what I am using it for it scales perfectly fine and at a fraction of the cost of the older brothers. Also, it should be noted that on a raspberry pi (at about $60) does the prediction with CPU at almost 200ms. 10ms vs 200ms... Nano wins!Update:If you are using Python I would suggest you buy the Xavier if you can afford it. If you are training models, this is not right either. I have a Xavier AGX for automated training and management of docker containers. I am moving to the Xavier NX for my evaluation needs as docker and python (tensorflow) take a LOT of RAM. This is a great buy for tinkering, but for heavy loads I would not bother. The great thing is that I learned a ton for less than $200.Also, make sure you update the resource usage after getting this live. This command sets it to max:sudo /usr/sbin/nvpmodel -m 0

Starting to show its' age in the fast paced A.I. compute hardware arena. Still a great value for about $100. Not the easiest SBC to use, but if you are familiar with Ubuntu Linux, you will not have any problems using it. There are a lot of helpful tutorials on the web, that can save a lot of time getting started. The Nvidia Jetson forum has a lot of helpful, searchable, information. Even boosted with the Maxwell GPU, this is not a "supercomputer" but it is very capable as a development system. You can use the Jetson Nano as a general purpose Linux computer, to run a small server, NAS, or firewall, but you can probably do that for even less money with a Raspberry Pi 4 or Intel Atom SBC. Jetson does not run x86 (Intel CPU) code. It uses a quad core ARM CPU, and Maxwell GPU.

1. My first Nano arrived broken and inoperable.2. My second Nano encountered errors in installation, which could have been addressed by some documentation, but it was not. I had to go searching online to fix an error that apparently happens with every installation of the Jetson Nano, but I was not warned about it in the documentation.3. If I turned the Nano off and then on again, it would often encounter software issues, and I would have to format and re-flash the disk, a process that takes up to an hour.4. No instructions are given on how to work with C files or Cmake.5. The (poor) documentation that comes with the Nano recommends following a link to a tutorial. There are 3 tutorials online, and this link in the documentation will send you to the least informative one. If you choose to buy this terrible, unreliable product, take the "intro to ai on the jetson nano" class.6. You speed up object detection by converting a tensorflow graph to tensorRT. There are no readable tutorials online for how to do this. Scan through Nvidia documentation and you will see what I mean: poorly written, convoluted, disorganized all over the place.7. When I ask for help on the Nvidia forums, I get less-than-helpful answers from Nvidia staff, such as "try this sample".8. There are few resources online for projects on the Nano and how to do them.9. The Board does not come with its own Wifi, the dongle must be purchased.10. Apparently the software that came with the Nano for remote desktop use does not even work.11. This product was lazily slapped together with little documentation, and no consideration for "makers" or beginners to the world of computing.12. I have spent 3 weeks trying to get this thing to do what I want it to do, consulted multiple people from Nvidia, and still no dice. I would have been done in 2 days if I had bought a different single board computer.

Informative ad, order was shipped promptly and delivered overseas in just a few days. Product well packed, complete and in perfect condition. Thank you for that! (Only one thing on the flipside: this is a hard-to-find-product, due to current shortage in the IC market, and I met demand-based pricing. That's why I have to give four stars on Value for money. Hopefully, market will get better, next year?).

After receiving my Jetson Nano I installed the Nano in a fan cooled chassis, burned a 64GB micro SD card with the Jetson Nano Ubuntu Mate OS, installed the micro SD card in the Nano, and proceeded to watch the Nano boot up per the Nvidia installation instructions. The OS found the Waveshare WiFi module that I had installed prior to boot up and allowed me to connect to my home WiFi router right away.I am using this Jetson Nano as a Ubuntu/ROS development system for communicating with my robot chassis that is running Ubuntu Xenial/ROS Kinetic. It is truly amazing how fast this Nano is compared to the Rpi 3B+s that I have been using.