Catalyst for the Intelligent Edge: A Comprehensive Analysis of the NVIDIA Jetson Nano—Architecture, Performance Benchmarking, and Comparative Standing in the AIoT Landscape

As the Internet of Things (IoT) increasingly merges with artificial intelligence, there is a growing demand for the efficient, positive, local and real-time processing of data at the edge of the network.This paper evaluates the NVIDIA Jetson Nano, an inexpensive, GPU-accelerated edge AI platform, by analyzing its architecture, benchmarking its performance on standardized AI workloads, and systematically comparing it with a selection of competitive edge AI boards (Raspberry Pi 4, Google Coral).Our results show Jetson Nano is an appealing point of entry for AI-enabled IoT applications such as smart surveillance, robotics, environmental monitoring, and health care, given the Jetson Nano's balance of price, energy consumption and software ecosystem.However, given its limited computational processing throughput and memory, the Jetson Nano is not suitable for more strenuous, multi-stream, or latency-sensitive IoT deployments.This paper provides developers with actionable implications for making selections between edge AI platforms of the IoT projects they are working on, and the paper identifies future work needed in hardware and software optimizations to unlock the potential power of intelligent edge systems.