This website might earn affiliate commissions from the hyperlinks on this web page. Phrases of use.

In the course of the Nvidia keynote at GTC right now, the corporate unveiled its new Jetson Orin Nano system-on-module (SOM) and introduced updates to its Nova Orin robotic reference platform.

At $199, the brand new Orin Nano is costlier than the older Xavier-based Jetson Nano. Nvidia claims that the brand new {hardware} is way sooner, and the system contains 4GB and 8GB storage choices, as an alternative of 2GB/4GB. Each tiers function an Ampere GPU and embrace deep studying and laptop imaginative and prescient accelerators. Nvidia says that the 8GB model can carry out “as much as 40 TOPS” for between seven and fifteen watts. The 4GB model ought to ship about half that efficiency, however with energy choices “as little as 5W to 10W.”

One important distinction between the AGX Orin product household (together with the Nano) and the earlier Nvidia Xavier SoC is their selection of CPU IP. Whereas each Xavier and Orin use ARM-based processors, Orin relies on the ARM Cortex-A78AE. Xavier’s CPU, in distinction, was codenamed Carmel. Carmel is an extra iteration of Nvidia’s long-running (and infrequently semi-mythical) “Mission Denver.” Initially envisioned as a CPU that may be able to emulating x86 and difficult Intel by itself turf, a few of Mission Denver’s design ideas had been much like Transmeta’s. Carmel was the second iteration of the core and confirmed important enhancements in a number of areas. It appears the core has been quietly retired in favor of an ARM-designed CPU implementation.

There might be good causes for this. The ARM Cortex-A78AE is designed to satisfy further security and reliability requirements that odd chips need not present, and it could be that Nvidia discovered it less complicated to construct round a regular ARM core than to proceed creating its personal structure.

Nvidia Proclaims 3 New Nova Orin Reference Platform Configurations

Orin modules specialise in coordinating knowledge from a wide selection of sensors, equivalent to IR cameras, ultrasonic sensors, and inertial measurement models. They’re able to wrangling a number of concurrent AI utility pipelines. Nvidia additionally emphasizes its Orin modules as a part of three new reference designs for autonomous cellular robots (AMRs). Along with the brand new Orin modules, Nvidia introduced an replace to its Nova Orin AMR reference platform. Or reasonably, a trio of updates. The corporate launched particulars for 3 totally different AMR configurations. All three depend on Jetson AGX Orin modules.

Nvidia Orin. Picture: Nvidia

Inward-facing sensors and cameras, in addition to AMRs utilizing the Nova Orin platform, can plug in to Nvidia’s Metropolis instruments, to supply each “inside-out” and “outside-in” mapping views. (Wall-E seems to be outward on the world from the within of his chassis, so his perspective is “inside-out.” The surveillance cameras in his Purchase-N-Giant warehouse look inward from the skin, in order that they’re “outside- in.”)

Nvidia’s Jetson AGX Orin modules run the corporate’s Isaac robotics stack, which Nvidia CEO Jensen Huang introduced right now can be shifting to the cloud. Isaac, together with Isaac Sim, proffers a collection of “cloud-native” instruments appropriate with the ROS 2 ecosystem. Builders can use Isaac Sim to handle sensor knowledge, which Nvidia hopes will discover widespread use in autonomous self-driving automobiles, immersive gaming, and medical diagnostics. For instance, SlashGear reviews that Amazon is utilizing Isaac Sim to map out the within of its warehouses, as a way to check, practice, and handle its personal AMRs.

GTC 2022 runs from Sept. 19 via Sept. 22.

Now Learn:

Subscribe Us to receive our latest news in your inbox!

We don’t spam! Read our privacy policy for more info.


Please enter your comment!
Please enter your name here