Back again in March, NVIDIA introduced Jetson Orin, the next-era of their ARM single-board computers supposed for edge computing apps. The new system promised to provide “server-course AI performance” on a board small adequate to set up in a robot or IoT device, with even the most affordable tier of Orin modules presenting roughly double the effectiveness of the prior Jetson Xavier modules. Regrettably, there was a bit of a catch — at the time, Orin was only offered in advancement package variety.
But now, NVIDIA has declared the quick availability of the Jetson AGX Orin 32GB production module for $999 USD. This is basically the mid-range offering of the Orin line, which tends to make releasing it initially a sensible plenty of alternative. People who need the best-close functionality of the 64GB variant will have to wait till November, but there’s still no difficult release day for the lesser NX Orin SO-DIMM modules.
Which is a bit of a letdown for people like us, since the two SO-DIMM modules are likely the most appealing for hackers and makers. At $399 and $599, their pricing will make them considerably additional palatable for the personal experimenter, when their lesser dimension and more acquainted interface should make them much easier to put into action into Do-it-yourself builds. Even though the Jetson Nano is nonetheless an unbeatable bargain for people looking to dip their toes into the CUDA waters, we could definitely see individuals investing in the significantly far more effective NX Orin boards for more elaborate initiatives.
When the AGX Orin modules may possibly be a bit steep for the regular tinkerer, their availability is still one thing to be thrilled about. Many thanks to the widespread JetPack SDK framework shared by the Jetson relatives of boards, applications developed for these greater-conclusion modules will mostly continue to be compatible across the complete product line. Confident, the cheaper and older Jetson boards will operate them slower, but as far as device discovering and AI purposes go, they’ll however run circles all around a thing like the Raspberry Pi.