NVIDIA has announced the Jetson Xavier NX, a 70 mm x 45 mm AI at the edge compute module that delivers up to 21 TOPS of performance and as little as 10 W of power consumption. It will be available in March 2020 for $399.

The module is mechanically and pin-compatible with the Jetson Nano, and includes:

• Compute: NVIDIA Volta GPU with 348 CUDA cores, 48 Tensor cores, two NVIDIA Deep Learning Accelerators (NVDLAs), and a 6-core Arm Carmel 64-bit CPU
• Memory: 6 MB L2 plus 4 MB L3 cache (on CPU) and 8GB 128-bit LPDDR4x DRAM
• Video: 2x 4K30 encode, 2x 4K60 decode
• I/O: GbE, PCIe, 12 MIPI CSI-2 lanes (3x4 or 6x2), I2Cs, GPIOs
• Software: Ubuntu-based Linux OS; NVIDIA JetPack SDK; and TensorFlow, PyTorch, MxNet, Caffe, and other AI frameworks

The Xavier NX is capable of running multiple neural networks in parallel and processing data from multiple high-resolution sensors, with support for up to six CSI cameras or 36 via virtual channels. The platform runs the CUDA-X AI software architecture, and delivers 51.2 GBps memory bandwidth for neural networking applications.

Developers interested in starting application development now can leverage the Jetson AGX Xavier Developer Kit and use a software patch to emulate the Jetson Xavier NX.

For more information, visit www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-xavier-nx.