Review on Studio Jetson Developer Computer Development by Alex Webster

When I first bought this I couldn't turn it on. I have poor eyesight and thought the jumper was in the correct position for the 5V 4A power supply. Be sure to pull it out and cover the other pin. Do not attempt to run it from a lower amperage power supply as performance will be halved or worse. Also make sure you have a fan. This thing gets hot and the GPU will slow down when it gets hot. The fans are difficult to screw onto the heatsink, but it's worth it as the performance increases by around 15-25%. Keep in mind that you must be using Nvidia's custom Ubuntu operating system. I tried a custom install but it didn't work. I use this for AI prediction. I started with 3 small networks. This takes about 10 ms per network. You wouldn't use this module for training, but for using a mockup it's amazing. Don't let that put you off if you're running small networks. If you're trying to capture full motion video and predict 60 times per second then get Xavier. If you do more intermittent detection, e.g. B. every 50 ms, this works. For what I use it scales beautifully and at a fraction of the cost of its bigger brothers. Also, it should be noted that on the Raspberry Pi (around $60) the prediction is done with the CPU at almost 200ms. 10 ms instead of 200 ms. Nano wins! Update: If you use Python I would advise you to buy Xavier if you can afford it. If you're training models, that's also wrong. I have Xavier AGX for automated training and docker container management. I'm switching to Xavier NX for my evaluation needs as Docker and Python (Tensorflow) take up a lot of RAM. It's a great buy for romping, but for heavy loads I wouldn't bother. The great thing is that I learned a lot for less than $200. Also, make sure to update your resource usage after you get this live. This command sets it to max:sudo /usr/sbin/nvpmodel -m 0