NVIDIA Jetson AGX Orin 32GB AI Computing Module
The NVIDIA Jetson AGX Orin 32GB module delivers up to 200 TOPS of AI performance with power configurable between 15W and 40W, providing more than 6X the performance of Jetson AGX Xavier in the same compact form factor. This powerful system-on-module (SOM) is designed for robotics, autonomous machines, edge AI applications, computer vision, and natural language processing across industries including manufacturing, logistics, healthcare, retail, and smart cities.
Next-Generation AI Performance for Edge Computing
Built on the NVIDIA Ampere architecture, the Jetson AGX Orin 32GB module features a 1792-core GPU with 56 tensor cores, an 8-core Arm Cortex-A78AE CPU, and 32GB of 256-bit LPDDR5 memory with 204.8 GB/s bandwidth. This system-on-module supports multiple concurrent AI application pipelines with next-generation deep learning accelerators (2x NVDLA v2.0), vision accelerators (PVA v2.0), high-speed I/O, and fast memory bandwidth for running the most complex AI models.
Key Features
-
200 TOPS AI Performance: Up to 108 INT8 Sparse TOPS or 54 FP16 TFLOPS for demanding AI inference workloads
-
NVIDIA Ampere GPU: 1792 CUDA cores and 56 3rd-generation tensor cores for accelerated computing
-
8-Core ARM CPU: Arm Cortex-A78AE v8.2 64-bit processor with 2MB L2 + 4MB L3 cache, up to 2.2 GHz
-
32GB LPDDR5 Memory: High-speed 256-bit memory with 204.8 GB/s bandwidth
-
64GB eMMC 5.1 Storage: Onboard storage for operating system and applications
-
Flexible Power Options: Configurable between 15W and 40W for optimal performance per watt
-
Compact Form Factor: 100mm × 87mm module with 699-pin connector and integrated thermal transfer plate
-
Multiple AI Accelerators: 2x NVDLA v2.0 deep learning accelerators and PVA v2.0 vision accelerator
Advanced AI Software Stack
Jetson AGX Orin runs the comprehensive NVIDIA AI software stack and supports industry-leading frameworks. Use-case specific application frameworks include NVIDIA Isaac for robotics, DeepStream for vision AI, and Riva for conversational AI. The module is compatible with JetPack SDK, which includes a complete Linux-based board support package, CUDA toolkit, cuDNN, TensorRT, and development tools for accelerated AI development.
Ideal Applications
The Jetson AGX Orin 32GB module excels in:
-
Robotics: Autonomous mobile robots (AMRs), collaborative robots, and warehouse automation
-
Computer Vision: Multi-camera processing, object detection, and real-time video analytics
-
Natural Language Processing: Conversational AI and speech recognition at the edge
-
3D Perception: Depth sensing and spatial AI for autonomous navigation
-
Multi-Sensor Fusion: Combining data from cameras, LiDAR, radar, and other sensors
-
Industrial AI: Quality inspection, predictive maintenance, and process optimization
-
Healthcare: Medical imaging analysis and diagnostic AI systems
-
Smart Cities: Traffic management and intelligent infrastructure
High-Speed Connectivity
The module provides extensive connectivity options through its 699-pin Molex Mirror Mezz connector, including:
- PCIe Gen4 interfaces (up to 8 lanes)
- USB 3.2 Gen2 (10 Gbps)
- Multi-Gigabit Ethernet (MGBE) up to 10 Gbps
- MIPI CSI-2 camera interfaces supporting up to 6 cameras (16 via virtual channels)
- Display outputs (HDMI 2.1, DisplayPort 1.4a)
- GPIO, I2C, SPI, UART, and CAN interfaces
Developer Ecosystem and Support
Access a broad ecosystem of partners offering AI and system software, developer tools, cameras, sensors, carrier boards, and custom development services. Accelerate development with NVIDIA Omniverse Replicator for synthetic data generation and NVIDIA TAO Toolkit for fine-tuning pre-trained AI models from the NGC catalog.
Technical Advantages
-
Form-Factor Compatible: Pin-compatible with Jetson AGX Xavier for easy upgrades
-
Efficient SOM Design: Integrated power management, memory, and thermal interface
-
Production Ready: Designed for deployment in commercial products
-
Scalable Platform: Part of the Jetson Orin family with multiple performance tiers
What's Included
- NVIDIA Jetson AGX Orin 32GB production module
- Integrated thermal transfer plate
- 699-pin board-to-board connector
Full Specifications
AI Performance
| AI Performance |
200 TOPS (INT8) |
| GPU Performance |
Up to 108 INT8 Sparse TOPS or 54 FP16 TFLOPS |
| CUDA Performance |
Up to 3.365 FP32 TFLOPS |
GPU
| Architecture |
NVIDIA Ampere |
| CUDA Cores |
1792 |
| Tensor Cores |
56 (3rd Generation) |
| Max GPU Frequency |
930 MHz |
CPU
| Processor |
8-core Arm Cortex-A78AE v8.2 64-bit |
| Cache |
2MB L2 + 4MB L3 |
| Max CPU Frequency |
2.2 GHz |
AI Accelerators
| Deep Learning Accelerator |
2x NVDLA v2.0 |
| DLA Max Frequency |
1.4 GHz |
| Vision Accelerator |
PVA v2.0 |
Memory
| System Memory |
32GB 256-bit LPDDR5 |
| Memory Bandwidth |
204.8 GB/s |
| Storage |
64GB eMMC 5.1 |
Video Processing
| Video Encode |
2x 4K60 | 4x 4K30 | 8x 1080p60 (H.265) |
| Video Decode |
2x 8K30 | 8x 4K30 | 18x 1080p60 (H.265) |
| JPEG Acceleration |
NVJPG for hardware-accelerated JPEG encoding/decoding |
Camera Support
| MIPI CSI-2 |
Up to 6 cameras (16 via virtual channels) |
| Maximum Bandwidth |
16 lanes MIPI CSI-2 |
Power
| Power Modes |
15W to 40W (configurable) |
| Input Voltage |
5V, 7V-20V |
Physical
| Dimensions |
100mm × 87mm |
| Connector |
699-pin Molex Mirror Mezz |
| Thermal Solution |
Integrated thermal transfer plate |
Software Support
| Operating System |
NVIDIA JetPack SDK (Linux-based) |
| AI Frameworks |
TensorRT, cuDNN, CUDA Toolkit |
| Application Frameworks |
Isaac (robotics), DeepStream (vision AI), Riva (conversational AI) |
| Development Tools |
Omniverse Replicator, TAO Toolkit, NGC Model Catalog |
Specifications are sourced from the official NVIDIA product documentation and may be updated. Please refer to the NVIDIA Jetson Download Center for the latest datasheets and technical documentation.
