NVIDIA Jetson AGX Orin 64GB AI Computing Module 275 TOPS Performance
The NVIDIA Jetson AGX Orin 64GB module delivers up to 275 TOPS of AI performance with power configurable between 15W and 60W, giving you more than 8X the performance of Jetson AGX Xavier in the same compact form factor. Designed for robotics, autonomous machines, edge AI, and vision AI applications, this advanced system-on-module (SOM) is the powerhouse behind next-generation intelligent systems across industries from manufacturing to smart cities, drones, and autonomous vehicles.
Why You'll Love the Jetson AGX Orin 64GB
-
Exceptional AI Performance: Up to 275 TOPS enables deployment of large, complex AI models for natural language understanding, 3D perception, and multi-sensor fusion
-
NVIDIA Ampere GPU Architecture: 2048 CUDA cores and 64 Tensor cores deliver advanced parallel computing with ray-tracing and TensorRT support
-
High-Speed Memory & Bandwidth: 64GB LPDDR5 unified memory with 204.8 GB/s bandwidth feeds multiple concurrent AI application pipelines seamlessly
-
Flexible Power Configuration: Scale performance between 15W and 60W to match your application's requirements
-
Compact, Credit-Card-Sized Form Factor: 100mm × 87mm module enables integration into space-constrained robotics and embedded designs
-
Comprehensive AI Software Stack: Runs JetPack SDK, NVIDIA Isaac for robotics, DeepStream for vision AI, and Riva for conversational AI
Unmatched AI Computing Power
At the heart of the Jetson AGX Orin 64GB is the NVIDIA Ampere architecture GPU with 2048 CUDA cores and 64 third-generation Tensor cores. This delivers up to 170 INT8 Sparse TOPS of Tensor compute and up to 5.3 FP32 TFLOPS of CUDA compute. The module also features dual NVDLA v2.0 deep learning accelerators and a Programmable Vision Accelerator (PVA) v2.0, enabling dedicated hardware acceleration for computer vision tasks like warping, image pyramids, filtering, and feature detection.
High-Performance CPU & Memory
The 12-core Arm Cortex-A78AE v8.2 64-bit CPU is arranged in three clusters of four cores each, with 2MB L3 cache per cluster and a 4MB system cache. Combined with 64GB of 256-bit LPDDR5 memory running at 204.8 GB/s bandwidth, the Jetson AGX Orin effortlessly handles demanding AI inference workloads, real-time vision processing, and multi-stream video analytics.
Built for Real-World AI Applications
Whether you're building autonomous mobile robots (AMR), industrial inspection systems, intelligent video analytics, drones, or medical imaging devices, the Jetson AGX Orin 64GB delivers the compute power and flexibility you need. It supports multiple concurrent AI pipelines for tasks like object detection, pose estimation, semantic segmentation, speech recognition, and natural language processing – all running simultaneously on a single device.
Developer-Friendly Ecosystem
The Jetson platform is supported by NVIDIA JetPack SDK, which includes a Linux-based board support package, CUDA toolkit, cuDNN, TensorRT, and a comprehensive set of computer vision and AI libraries. Use NVIDIA TAO Toolkit to fine-tune pretrained AI models from the NGC catalog, and leverage Omniverse Replicator for synthetic data generation. Ecosystem partners provide additional software, cameras, sensors, carrier boards, and custom design services to accelerate your product development.
High-Speed I/O & Connectivity
The module exposes a rich set of interfaces through a 699-pin board-to-board connector, including:
-
PCIe Gen 4: 7 controllers with 22 shared lanes, up to 16 Gbps per lane
-
USB 3.2 Gen 2: Up to 3 ports at 10 Gbps, plus USB 2.0 support
-
High-speed Ethernet: Multi-gigabit Ethernet (MGBE) support
-
Camera Interfaces: Multiple MIPI CSI-2 camera inputs for vision AI applications
-
Display Outputs: DisplayPort and HDMI support for visualization and debugging
-
Serial Interfaces: UART, SPI, I2C, I2S for sensors and peripherals
Form Factor & Pin Compatibility
The Jetson AGX Orin 64GB SOM measures 100mm × 87mm and is form-factor and pin-compatible with Jetson AGX Xavier, enabling easy migration and reuse of existing carrier board designs. The standardized thermal and mechanical interfaces simplify integration and protect the module from external mechanical forces.
What's in the Box
- NVIDIA Jetson AGX Orin 64GB System-on-Module (SOM)
- Quick start documentation (digital)
Full Specifications
GPU
| Architecture |
NVIDIA Ampere with ray-tracing cores |
| CUDA Cores |
2048 |
| Tensor Cores |
64 (3rd generation) |
| GPU Max Frequency |
Variable based on power mode |
| AI Performance |
Up to 275 TOPS (INT8 Sparse with all accelerators), up to 170 INT8 Sparse TOPS (Tensor cores), up to 5.3 FP32 TFLOPS (CUDA) |
CPU
| CPU |
12-core Arm Cortex-A78AE v8.2 64-bit |
| CPU Clusters |
3 clusters of 4 cores each |
| L3 Cache |
2MB per cluster |
| System Cache |
4MB |
Memory
| Memory |
64GB 256-bit LPDDR5 |
| Memory Bandwidth |
204.8 GB/s |
| Storage |
64GB eMMC 5.1 |
AI Accelerators
| Deep Learning Accelerator |
2x NVDLA v2.0 |
| Vision Accelerator |
PVA v2.0 (Programmable Vision Accelerator) |
Video Encode/Decode
| Video Encoder |
2x 4K60 (H.265) | 6x 4K30 (H.265) | 14x 1080p60 (H.265) |
| Video Decoder |
2x 8K30 (H.265) | 6x 4K60 (H.265) | 16x 1080p60 (H.265) |
Connectivity & I/O
| PCIe |
7 controllers, 22 shared lanes, PCIe Gen 4 (up to 16 Gbps/lane) |
| USB |
Up to 3x USB 3.2 Gen 2 (10 Gbps), up to 4x USB 2.0, 1x USB 3.2 Gen 1 device mode (5 Gbps) |
| Ethernet |
Multi-gigabit Ethernet (MGBE) support |
| Camera Interfaces |
Multiple MIPI CSI-2 (Virtual channels supported) |
| Display |
DisplayPort, HDMI (via carrier board implementation) |
| Serial/Other |
4x UART, 3x SPI, 8x I2C, 4x I2S/DAP audio ports, 2x PDM |
Power
| Power Modes |
Configurable between 15W and 60W |
| Maximum Performance Mode |
MAXN mode up to 60W |
| Power Inputs |
Three power inputs via module connector |
Mechanical
| Form Factor |
System-on-Module (SOM) |
| Dimensions |
100mm × 87mm |
| Connector |
699-pin board-to-board connector |
| Compatibility |
Form-factor and pin-compatible with Jetson AGX Xavier |
Software
| SDK |
NVIDIA JetPack SDK (includes Linux BSP, CUDA, cuDNN, TensorRT) |
| OS |
Linux (Ubuntu-based) |
| AI Frameworks |
NVIDIA Isaac (robotics), DeepStream (vision AI), Riva (conversational AI), TAO Toolkit, Omniverse Replicator |
Specifications are sourced from the manufacturer product page and may be updated.
