NVIDIA has just announced Jetson Xavier NX system-on-module, with the company claiming it is the “world’s smallest, most powerful AI supercomputer for robotic and embedded computing devices at the edge” with a 70x45mm “Jetson Nano” form factor, and delivering either up to 14 TOPS at 10 Watts or 21 TOPS at 15 Watts.
The company expects the module to be used in small commercial robots, drones, intelligent high-resolution sensors for factory logistics and production lines, optical inspection, network video recorders, portable medical devices, and other industrial IoT systems.
- SoC – NVIDIA Xavier with 6-core NVIDIA Carmel ARM v8.2 64-bit CPU,
6MB L2 + 4MB L3 caches, and a 384-core NVIDIA Volta GPU with 48 Tensor Cores, 2x NVDLA deep learning accelerators delivering up to 21 TOPS at 15 Watts
- System Memory – 8 GB 128-bit LPDDR4x @ 51.2GB/s
- Storage – 16 GB eMMC 5.1 flash
- 2x 4K @ 30 (HEVC)
- 6x 1080p @ 60 (HEVC)
- 2x 4K @ 60 (HEVC)
- 12x 1080p @ 60 (HEVC)
- 32x 1080p @ 30 (HEVC)
- 260-pin SO-DIMM connector exposing:
- Display – 2x multi-mode DP 1.4/eDP 1.4/HDMI 2.0
- Networking – 10/100/1000 Mbps Ethernet
- Up to 6 CSI Camera (36 via virtual channels), 12 lanes MIPI CSI-2, D-PHY 1.2 up to 30 Gbps
- 1x PCIe x1, 1x PCIe x4 (PCIe Gen3, Root Port & Endpoint)
- I2C and GPIO’s
- Power – Configurable 10W or 15W
- Dimensions – 69.6 x 45 mm
Like other Jetson boards and modules, Jetson Xavier NX runs Ubuntu-based Linux, is supported by NVIDIA JetPack software development kit, and CUDA-X AI libraries. The module supports all most common AI frameworks such as TensorFlow, PyTorch, MxNet, Caffe, and others.
There’s no Jetson Xavier NX development kit, but since the SoM is pin-compatible with Jetson Nano, so you should be able to reuse/upgrade existing carrier boards. In order to get started quickly, the company invites developers to begin application development using the Jetson AGX Xavier Developer Kit with a software patch emulating Jetson Xavier NX.
NVIDIA Jetson Xavier NX SoM will be available in March for $399 to companies looking to create high-volume production edge systems, which reads like it won’t sell to individuals like makers or students. More details may be found in the product page.