The NX ('Nano Xavier') is said to deliver up 21 TOPS of AI performance, yet only consumes 15W. That figure drops to 14 TOPS in 10W mode.
The NX uses Nvidia's Volta processor containing 384 Cuda cores, 48 tensor cores, and two deep learning accelerators, plus a six-core Carmel Arm CPU.
It can run all Cuda-X models, and new software can be developed for the device by using a patched copy of the Jetson AGX Xavier developer kit.
The NX is said to be suitable for applications including autonomous machines (eg, self-navigating drones), high-resolution sensors (eg, for automated product inspections), video analytics (eg, to improve the analytics performance of network video recorders without increasing their size), and embedded devices (eg, medical imaging and DNA sequencing).
"AI has become the enabling technology for modern robotics and embedded devices that will transform industries," said vice president and general manager of edge computing Deepu Talla.
"Many of these devices, based on small form factors and lower power, were constrained from adding more AI features. Jetson Xavier NX lets our customers and partners dramatically increase AI capabilities without increasing the size or power consumption of the device."
Delivery of the US$399 device is scheduled to start in March 2020.
In related news, Nvidia has claimed top spot (on a per-processor basis) in all of the latest MLPerf Inference 0.5 benchmarks.
It was the only vendor to submit results for every test.
"AI inference is harder than it seems," said product management director Paresh Kharya. The results show that "CPUs are no longer good enough for AI inference."
Nvidia's product range can cope with increasingly diverse and complex neural nets, and the extensive pre- and post-processing required in some applications.
Additionally, it offers a single hardware and software architecture across cloud, data centre, edge, and mobile devices.
A recent win for the company was the US Postal Service's adoption of Nvidia's platform for a system that will read the labels on pieces of mail and sort them accordingly, eg by the presence of hazardous materials.
"AI is at a tipping point as it moves swiftly from research to large-scale deployment for real applications," said Nvidia general manager and vice president of accelerated computing Ian Buck.
"AI inference is a tremendous computational challenge. Combining the industry's most advanced programmable accelerator, the Cuda-X suite of AI algorithms and our deep expertise in AI computing, Nvidia can help data centres deploy their large and growing body of complex AI models."