New module targets demand for real-time processing capabilities in AI

18-10-2022 | Impulse Embedded | Industrial

NVIDIA has announced the Jetson Orin Nano to be released in Q1 2023 and will be integrated into various rugged Edge Computers from Impulse Embedded. Sitting on the entry-level side of Jetson’s offerings, this new module targets the growing demand for real-time processing capabilities in AI devices such as smart cameras, smart meters, handhelds, and service robots, where integrators often need lower power and a more cost-effective solution.

The new module provides up to 80x the AI performance of its predecessor, supplying a substantial boost in on-device, low-latency processing with no power consumption or cost increase.

The series will be offered in two versions, a 4GB and an 8GB model, each denoting the amount of LPDDR5. The 4GB version offers a 512-core NVIDIA Ampere GPU with 16 Tensor cores, while the 8GB version provides a 1024-core NVIDIA Ampere GPU with 32 Tensor cores. Each version shares the same MAX GPU frequency of 625MHz, delivering 40TOPS.

The versions will include a six-core Arm Cortex-A78AE CPU, Ampere architecture GPU, video decode engine, ISP, audio processing engine, video image compositor, and video input block.

The modules also offer various high-speed interfaces, including up to seven lanes of PCIe Gen3, 3x 10Gbps, USB3.2 Gen 2, eight MIPI CSI-2 camera ports, and a comprehensive range of supported sensor I/O, further lending these modules towards those lightweight applications discussed previously.

Those familiar with the Jetson range may be aware that NVIDIA has always set out with the idea of scalability, cost-effectiveness and simplicity. Therefore, it features the same 70mm x 45mm 260-pin SODIMM footprint found in the Nano, TX2 NX, Xavier NX and Orin NX modules, which can lower engineering costs when scaling up.

Impulse Embedded can deliver fully deployable embedded Edge AI computing solutions straight out of the box.

By Seb Springall