⚡Nvidia Jetson Orin Nano 2 Doubles Edge AI Speed at 15W
TL;DR
Nvidia's new Jetson Orin Nano 2 packs 78 TOPS and 8GB into the same board as the Orin Nano Super, at twice the inference throughput. In 15-watt mode it draws 40% less power for the same work.
Nvidia's new Jetson Orin Nano 2 packs 78 TOPS and 8GB into the same board as the Orin Nano Super, at twice the inference throughput. In 15-watt mode it draws 40% less power for the same work. The pitch: small models are finally accurate enough to run the robot instead of the data center.
Key Points
78 trillion operations per second of AI compute, 8GB memory, eight-core Arm CPU
2x the inference performance of Jetson Orin Nano Super, credited to new Tensor Cores and higher memory bandwidth
At 15W it matches Orin Nano Super performance while using 40% less power
Deepu Talla, VP of Robotics and Edge AI: small and medium models now hit accuracy only frontier models reached a year ago
Runs LLMs and vision-language models locally via Nvidia's open Jetson software stack and agent skills
Why It Matters
If a $250-class board can run a VLM in real time at 15 watts, the cost floor for autonomous drones and inspection robots drops hard, and cloud round-trips stop being the default design.
Quick Facts
Frequently Asked Questions
Why does this matter?
If a $250-class board can run a VLM in real time at 15 watts, the cost floor for autonomous drones and inspection robots drops hard, and cloud round-trips stop being the default design.
What happened?
Nvidia's new Jetson Orin Nano 2 packs 78 TOPS and 8GB into the same board as the Orin Nano Super, at twice the inference throughput. In 15-watt mode it draws 40% less power for the same work.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,317 builders reading daily.