At GTC 2023 today, NVIDIA Corp. announced the NVIDIA Jetson Orin Nano Developer Kit, which it said “sets a new standard for creating entry-level AI-powered robots, smart drones, and intelligent vision systems.” It is also designed to simplify getting started with the NVIDIA Jetson Orin Nano series.
“Compact design, numerous connectors, and up to 40 TOPS [trillions of operations per second] of AI performance make this developer kit ideal for transforming visionary concepts into reality,” claimed NVIDIA.
The developer kit consists of a Jetson Orin Nano 8 GB module and a reference carrier board that can accommodate all NVIDIA Jetson Orin Nano and NX modules, providing a platform for prototyping next-generation edge AI products.
The Jetson Orin Nano 8 GB module features an NVIDIA Ampere architecture GPU with 1024 CUDA cores, 32 third-generation Tensor Cores, and a six-core Arm CPU. They are designed to enable multiple concurrent AI application pipelines and high-performance inference, said the company.
The developer kit carrier board boasts a wide array of connectors, including two MIPI CSI connectors supporting camera modules with up to four lanes, for higher resolution and frame rates than before, according to NVIDIA.
The company said the prior-generation Jetson Nano Developer Kit made AI accessible to everyone. It added that the new Jetson Orin Nano Developer Kit offers 80x the performance, enabling developers to run modern AI models, including transformer and advanced robotics models.
Not only does the kit provide a significant boost in AI performance over the prior-generation Jetson Nano, said NVIDIA, but it also provides 5.4x the CUDA compute, 6.6x the CPU performance, and 50x the performance per watt.
NVIDIA Jetson Orin Nano Developer Kit features
The NVIDIA Jetson Orin Nano Developer Kit includes an Orin Nano 8 GB module with an SD-card slot, a reference carrier board, preassembled heatsink/fan, a 19V DC power supply, and an M.2-Key E based wireless networking module.
Table 1: NVIDIA Jetson Orin Nano Developer Kit module specifications
|Module||NVIDIA Jeton Orin Nano 8 GB Module|
|GPU||NVIDIA Ampere architecture with 1024 NVIDIA CUDA Cores and
32 Tensor Cores
|CPU||6-core Arm Cortex-A78AE v8.2 64-bit CPU 1.5 MB L2 + 4 MB L3|
|Memory||8 GB 128-bit LPDDR5 68 GB/s|
|Storage||External through microSD slot External NVMe through M.2 Key M|
|Power||7 W to 15 W|
In addition to the bootable microSD card slot, NVIDIA provided two M.2 Key-M NVMe sockets are provided on the underside of the carrier for high-speed storage.
Table 2: NVIDIA Jetson Orin Nano Developer Kit reference carrier board specifications
|Camera||2x MIPI CSI-2 22-pin Camera Connectors|
|M.2 Key M||x4 PCIe Gen3|
|M.2 Key M||x2 PCIe Gen3|
|M.2 Key E||PCIe (x1), USB 2.0, UART, I2S, and I2C|
|USB||Type A: 4x USB 3.2 Gen2 Type C: 1x for Debug and Device Mode|
|Networking||1x GbE Connector|
|Display||DisplayPort 1.2 (+MST)|
|microSD slot||UHS-1 cards up to SDR104 mode (under the Module)|
40-pin Expansion Header (UART, SPI, I2S, I2C, GPIO)
100 x 79 x 21 mm (3.9 x 3.1 x 0.8 in.)
The Jetson Orin Nano Developer Kit is now available for preorder at $499, and will start shipping in April. To access the Jetson Orin Nano technical documentation, reference design files, and software, visit the Jetson Download Center.
Jetson Orin Nano runs all modern AI models
The Jetson Orin Nano Developer Kit, with up to 40 TOPS of AI performance, can run all modern AI models, asserted NVIDIA. This compute makes demanding AI applications possible, including running transformer models right at the edge, which was not possible before with Jetson Nano, it said.
Transformer models are the basis of recent generative AI applications like ChatGPT and DALL-E. A transformer model learns context and meaning by tracking the relationship between elements in sequential data, eliminating the need for a large labeled dataset, explained NVIDIA.
Jetson Orin Nano includes support for:
- City Segmentation for segmenting urban cityscapes into different classes
- PeopleNet Transformer based on the Deformable Detection Transformer
- BERT for natural language processing (NLP)
Jetson software accelerates AI and TTM
Jetson Orin Nano Developer Kit runs the NVIDIA AI software stack, with available use-case-specific application frameworks. These include NVIDIA Isaac for robotics, NVIDIA DeepStream for vision AI, and NVIDIA Riva for conversational AI.
Developers can save time with NVIDIA Omniverse Replicator for synthetic data generation (SDG), and with the TAO Toolkit for fine-tuning pretrained AI models from the NGC catalog, said NVIDIA.
Here are results from running some computer vision benchmarks with Jetson Orin Nano using the latest NVIDIA JetPack 5.1.1. These results show that the developer kit raises the bar for entry-level computer vision.
Testing included some dense INT8 and FP16 pretrained models from NGC, and an Industry Resnet-50 Benchmark.
The benchmark testing included the following:
- NVIDIA PeopleNet v2.5 for the highest accuracy people detection
- NVIDIA ActionRecognitionNet 2D and 3D models
- NVIDIA LPRNet for license plate recognition
- NVIDIA DashCamNet, BodyPoseNet for multiperson human pose estimation
- ResNet-50 (224×224) model for object detection
The Jetson Orin Nano Developer Kit can support models trained with NVIDIA TAO Toolkit 4.0, and it will soon add support for newly announced models in TAO Toolkit 5.0, said NVIDIA.
With TAO Toolkit 5.0, developers can take advantage of vision transformer models for image classification, object detection, and segmentation use cases. To learn more, see “Access the Latest in Vision AI Model Development Workflows with NVIDIA TAO Toolkit 5.0.”
The combination of NVIDIA Jetson Orin Nano and NVIDIA DeepStream is also suitable for edge applications, such as smart retail, smart city intersections, and industrial automation, said NVIDIA. The upcoming version of DeepStream, along with the introduction of the GXF runtime, allows Jetson Orin Nano to run AI graphs that require tight integration with deterministic systems, common in factory automation use cases, it added.
In addition, users can quickly learn to use DeepStream by building applications with the latest version of DeepStream Graph Composer. They can then deploy them to a Jetson Orin Nano with the click of a button, said NVIDIA.
NVIDIA Isaac on NVIDIA Jetson Orin offers to accelerate robot app development
NVIDIA said its Isaac robotics platform gives developers a platform for the development, simulation, and deployment of AI-enabled robots. Specifically, NVIDIA Isaac ROS, a collection of hardware-accelerated packages, makes it easier for ROS 2 developers to build high-performance solutions on the Jetson Orin Nano Developer Kit, said the company.
The new NVIDIA Isaac ROS DP release optimizes ROS 2 nodes processing pipelines, which can be executed on the Jetson Orin Platform. It also provides new DNN-based GEMS designed to increase throughput.
The image below shows the results of running these robotics packages on the Jetson Orin Nano using the upcoming NVIDIA Isaac ROS DP3 release. The performance is measured under load including message transport costs in RCL for practical benchmarking indicative of real-world performance.
Testing included the following algorithms:
- Visual SLAM that enables a robot to compute its location and movement from images by tracking visual features around its environment
- April Tags for AprilTag detection and pose estimation
- Image Detection
- Image Segmentation
- Proximity Segmentation to determine whether an obstacle is within a proximity field and to avoid collisions with obstacles during navigation
- Stereo Disparity for taking stereo input images and generating a disparity map of the input image for robot navigation
Get started with the NVIDIA Jetson Orin Nano Developer Kit
NVIDIA said developers can start prototyping their next-generation robotics and AI applications by preordering the NVIDIA Jetson Orin Nano Developer Kit and installing the latest NVIDIA JetPack.
The full lineup of the NVIDIA Jetson Orin family of production modules is available now, including the Jetson AGX Orin Series, the Jetson Orin NX Series, and the Jetson Orin Nano Series. The Jetson AGX Orin Developer Kit has also now been upgraded to include 64 GB of memory for the same price of $1,999.
For more information about the NVIDIA Jetson Orin Nano Developer Kit, watch the GTC session, “Connect with the Experts: A Deep-Dive Q&A with Jetson Embedded Platform Engineers.” To learn more about the module, see “Solving Entry-Level Edge AI Challenges with NVIDIA Jetson Orin Nano” and the Jetson Orin Nano Series Data Sheet.
The NVIDIA Embedded Developer page and forums allow NVIDIA users to find help from community experts.