NVidia has been manufacturing artificial intelligence-focused single board computers (SBCs) for years. They were powerful, however, they weren’t cheap. The Jetson TX1 and TX2 were both over $400 dollars and equipped for computer vision applications (for example: autonomous cars).
One of the most popular single board computers out there right now is the Raspberry Pi, which is more focused on offering a low price point for people who want to learn how to program, and for other educational purposes. $35 is the price for their most powerful model (Raspberry Pi 3B+), while some models cost $25 (Pi 3A+) and even as little as $5 (Pi Zero).
The Raspberry Pi’s, however, are not AI-focused or built for advanced AI applications such as autonomous cars. However the Raspberry Pi Compute module is an entry into the industrial sector, offering an SBC in a compact SODIMM module. The NVidia Jetson Nano is now offered as a $99 evelopment kit for those looking to develop A.I.-related apps on the Jetson platform, putting it well within reach for many people.
The NVidia Jetson Nano can run Linux, has 4GB of 64-bit LPDDR4 RAM, gigabit Ethernet, and is powered by an ARM A57 processor and a Maxwell-based 128-Cuda core GPU that NVidia says is capable of 472 gigaflops. The compact 70 x 45mm compute module has 16GB of eMMC flash memory for storage and consumes an impressively low 5 to 10 Watts.
This low power usage, combined with its ability to run multiple neural networks in parallel and utilize several high-resolution sensors simultaneously makes it enticing for ambitious DIY robotics projects. The low power consumption also makes it feasible for space-constrained, portable battery-powered devices.
For comparison, the bigger Jetson TX2 has 8GB of 128-bit LPDDR4 RAM and 32GB of eMMC flash memory, among other advantages. The Jetson Nano kit is $99, however, there will be a production-ready version for $129. The Nano kit is now available for pre-order and is expected to start shipping at the end of March.