NVIDIA Jetson AGX Orin 32GB production module is now in mass production and available after the 12-core Cortex-A78E system-on-module was first announced in November 2021, and the Jetson AGX Orin developer kit was launched last March for close to $2,000.
Capable of up to 200 TOPS of AI inference performance, or up to 6 times faster than the Jetson AGX, the NVIDIA Jetson AGX Orin 32GB can be used for AI, IoT, embedded, and robotics deployments, and NVIDIA says nearly three dozen partners are offering commercially available products based on the new module.
Here’s a reminder of NVIDIA Jetson AGX Orin 32GB specifications:
- CPU – 8-core Arm Cortex-A78AE v8.2 64-bit processor with 2MB L2 + 4MB L3 cache
- GPU / AI accelerators
- NVIDIA Ampere architecture with 1792 NVIDIA CUDA cores and 56 Tensor Cores @ 1 GHz
- DL Accelerator – 2x NVDLA v2.0
- Vision Accelerator – PVA v2.0 (Programmable Vision Accelerator)
- AI Performance – 200 TOPS (INT8) @ 50W / 100 TOPS for “dense” inteference
- Video Encode – 1x 4K60 | 3x 4K30 | 6x 1080p60 | 12x 1080p30 (H.265)
- Video Decode – 1x 8K30 | 2x 4K60 | 4x 4K30 | 9x 1080p60| 18x 1080p30 (H.265)
- System Memory – 32GB 256-bit LPDDR5 @ 204.8 GB/s
- Storage – 64GB eMMC 5.1 flash
- 699-pin Molex Mirror Mezz connector with
- Storage – Single lane UFS
- Display – 1x 8K60 multi-mode DP 1.4a (+MST), eDP 1.4a, HDMI 2.1
- Up to 6x CSI Camera (16 via virtual channels)
- 16 lanes MIPI CSI-2
- D-PHY 1.2 (up to 40Gbps) | C-PHY 1.1 (up to 164Gbps)
- Networking – 1x GbE, 1x 10GbE
- PCIe – 2x PCIe x8 (or 1x PCIe x8 + 2x PCIe x4), 1x PCIe x4, 2x PCIe x1 (PCIe Gen4, Root Port & Endpoint)
- USB – 3x USB 3.2, 4x USB 2.0
- Low-speed IOs – 4x UART, 3x SPI, 4x I2S, 8x I2C, 2x CAN, DMIC & DSPK, GPIOs
- Power Modes – 15W, 30W, or 50W
- Dimensions – 100 x 87mm
The Jetson AGX Orin 64GB delivering 275 TOPS and found in the developer kit is not available yet, but the production module is scheduled for November, and the lower costs Jetson Orin NX modules with 8GB or 16GB RAM will start selling “later this year”.
On the software side, the Jetson AGX Orin modules are supported by the usual NVIDIA Jetson software stack, which includes the NVIDIA CUDA-X accelerated stack, as well as Isaac for robotics, DeepStream for computer vision, Riva for natural language understanding, TAO Toolkit to accelerate model development with pre-trained models, and the Metropolis suite that combines “an application framework, a set of developer tools and partner ecosystem that brings visual data and AI together to improve operational efficiency and safety across industries”.
We’ll cover some third-party carrier boards and full hardware systems based on Jetson AGX Orin as they are announced, but NVIDIA tells us AAEON, Auvidea, Connect Tech, MiiVii, Plink, Realtimes, and TZTEK will shortly launch such hardware. Over 350 camera and sensor options are also available from Allied Vision, Appropho, Basler AG, e-Con Systems, Framos, Leopard Imaging, LIPS, Robosense, Shenzhen Sensing, Stereolabs, Thundersoft, Unicorecomm, and Velodyne.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.