Nvidia offers a small piece of hardware to power the AI projects of "developers, makers and enthusiasts"-- the Jetson Nano, a board promising to deliver 472 GFLOPS of processing power from a package measuring just 70 x 45mm.
The Nano is the latest offering in the Jetson line of embedded computing boards aimed at edge computing applications. What makes the Nano unique is the pricetag, since it costs all of $99 while still packing some fairly impressive specifications. For the curious, it carries a quad-core ARM A57 processor running at 1.43GHz, a 128-CUDA core Maxwell GPU, 16GB eMMC flash and 4GB DDR4 RAM, all within a 5W power envelope.
Nvidia says the Nano runs a range of AI frameworks, including TensorFlow, PyTorch, Caffe, Keras, and MXNet, while connectivity includes both USB-A and B, ethernet, DisplayPort, HDMI and even M.2. Users can load the OS, such as a Linux variant, via micro SD card, and power comes through micro USB cabel (with a barrel connector allowing for more powerful PSUs).
For the enthusiast wanting to get started in AI Nvidia offers a suitable add-on-- the JetBot, an autonomous robitics kit consisting of a chassis, battery pack and motors. Once put together the kit forms a self-driving robot users can control using collision-avoidance algorithms and AI such as OpenPose and Tiny Yolo.
The Jetson Nano should be available by end March 2019.