NVIDIA Jetson AGX Orin Boosts Edge AI Computing By 6X

NVIDIA just announced the Jetson AGX Orin, its latest Jetson embedded computer. The Jetson family is designed to be integrated into drones, industrial robots, and more. In short, they are the NVIDIA computing platform that’s on the frontlines and at the edge of our communications networks.

The specifications show that it has six times the computing power of its predecessor (200 trillion operations per second vs. 32). NVIDIA has not announced the Jetson AGX Orin price yet, but the previous-generation Jetson AGX Xavier costs $1000-$1500. Let’s wait and see.

Jetson AGX Orin is based on the NVIDIA Ampere architecture (2020), a generational leap over the NVIDIA Volta architecture (2013) used in the Jetson AGX Xavier. Ampere is the base architecture of the RTX 3000 series of graphics processors known to gamers worldwide.

Concretely, Jetson AGX Orin has enough power to run newly announced technologies such as Omniverse Avatar, a platform to generate and customize customer-facing avatars.

These virtual assistants would be completely autonomous and driven by machine learning specific to their industry. Omniverse Avatar essentially allows developers to focus on the value of the interaction rather than implementing their avatar.

Since Jetson computers can be integrated into vehicles, NVIDIA also presented an Omniverse demo. A driver has a natural voice interaction with the onboard computer (represented by an avatar) and asks for a notification that would usually require complex touchscreen inputs, thus increasing safety.

Based on the same architecture as the Jetson AGX Orin, the Drive AGX Orin platform runs the newly announced NVIDIA Concierge and Chauffeur software. Drive Chauffeur drives from address to address as shown in the video below. I don’t expect it to reach level-5 in the short term, but it would be interesting to see if NVIDIA has a timeline on that.

Concierge helps vehicles perform very complex tasks like Parking, with the ultimate goal of having the driver get out of the car and leave it to the AI to find a suitable parking spot. Yes! Sign me in.

We haven’t scratched the surface of what’s possible with having such hardware platforms in everyday life/work machinery and in many ways, robotics might be holding back many markets.

What’s remarkable is that NVIDIA is accelerating software development even faster than hardware computation speeds. And there lies NVIDIA’s moat against competitors that tend to (barely) compete on hardware only.

The platforms introduced today save incredible amounts of development and testing time to developers, and that’s worth serious money. In the end, they allow entire industries to explore new ways to benefit from the unfolding simulation and automation revolution.

Jetson AGX Orin isn’t yet available for purchase, but developers can register on its official homepage to be notified when units will be ready for shipping.

Hubert Nguyen
Categories: Computers, Events, Robots
Tags: AI (Artificial Intelligence), GPU (Graphics Processing Unit), gtc, NVIDIA, Self-Driving Cars