NVIDIA’s Jetson AGX Orin Development Kit is a seamless embedded platform to deliver 275 TOPS of pure performance. It targets programs like robotics, but it also works well for any embedded application that requires synthetic intelligence (AI) and device learning (ML) in the field.
The Jetson AGX Orin module (see figure) is built around a dozen 64-bit Arm Cortex-A78AE v8. 2 cores that have access to a 3 MB L2 cache and a 6 MB L3 cache, as well as a GPU based on the new NVIDIA processor. includes 2048 NVIDIA CUDA cores and 64 Tensor cores plus a pair of NVIDIA Deep Learning (NDLA) v2. 0 accelerators for AI/ML applications. the platform.
The hardware is impressive, but it’s NVIDIA’s software that encompasses it that really makes the difference. The Jetson Jetpack SDK is built on Ubuntu Linux, adding a UEFI and OP-TEE-based bootloader as a reliable runtime environment. The progression kit comes with everything installed, so it’s much less difficult to get started. Includes for the underlying equipment CUDA, CuDNN, TensorRT and DeepStream.
Two items that stand out in the newest kit are NVIDIA Train-Adapt-Optimize (TAO) and NVIDIA RIVA. The latter targets voice AI, while TAO is based on pre-trained models from NVIDIA. These can be custom designed in Microsoft’s Azure cloud. .