Yesterday NVidia announced the Jetson X1. It is an exceptionally powerful development board, with a specialized graphics processing unit that makes quick work of complex math problems. It’s also quite expensive, retailing at $599 (or $299 with an educator’s discount). We don’t think it’s a board you’ll want to run out and buy right now for your next generic project, but if you’re building an autonomous vehicle, you might want to check it out.
The specs of the X1 are impressive:
- GPU: 1 teraflops, 256-core Maxwell architecture-based GPU
- CPU: 64-bit ARM A57 CPUs
- Video: 4K video encode and decode
- Memory: 4GB LPDDR4; 25.6 gigabits/second
- Storage: 16GB eMMC
- Wi-Fi/Bluetooth: 802.11ac 2×2 Bluetooth ready
- Networking: 1GB Ethernet
- Size: 50mm×87mm
- Amazingly, this board uses only 10W of power
The board is as small as a credit card and, NVidia says, is ready for deployment in products. Applications, according to NVidia, include data centers involved in machine learning, autonomous vehicle research, and other specialized industrial uses. All of these types of applications require realtime computation of enormous data sets. An autonomous car must acquire millions of data points and evaluate if the numbers imply an open road or someone crossing the street.
Though details are sparse on the companies using the X1 in their products, we do know that the robot Jibo is powered by one. And the new DJI drone platform Manifold uses the beefy processing power of the X1 to target high-end drone design.
The X1 is also a candidate as a main board for a desktop Linux system. With optimized and accelerated video drivers, a bespoke version of Ubuntu Linux, and plenty of ways to connect to the internet, it is a decent value for the power it contains. Yes, the board renders Doom 3 graphics quite beautifully. But, as we have said, as a development board to do anything less than compute intensive applications, it is overkill.