NVIDIA Jetson AGX Orin 64GB LPDDR5 Module for Embedded AI

NVIDIA Jetson AGX Orin 64GB Module

The NVIDIA Jetson AGX Orin™ 64GB module delivers up to 275 TOPS of AI performance in a compact, production-ready form factor, with power configurable between 15W and 60W. Offering more than 8× the performance of Jetson AGX Xavier™ at similar size, it’s designed for advanced robotics, autonomous machines, and complex edge AI systems where performance and efficiency are critical.

Next-Generation GPU and Accelerators

Built on the NVIDIA Ampere GPU architecture, the module combines 2048 CUDA cores and 64 Tensor Cores with next-generation deep learning and vision accelerators. This heterogeneous compute architecture enables multiple concurrent AI pipelines, making it ideal for demanding tasks like natural language understanding, 3D perception, and multi-sensor fusion. A 12-core Arm Cortex-A78AE CPU with 3MB L2 and 6MB L3 cache further enhances multi-threaded and real-time workloads.

High-Speed Memory and Storage

Jetson AGX Orin 64GB is equipped with 64GB of LPDDR5 memory on a 256-bit interface, delivering 205 GB/s bandwidth to feed the most demanding AI models. Onboard storage includes 64GB of eMMC 5.1, with additional expansion possible through external storage solutions provided by carrier boards.

Rich Multimedia and Camera Support

Developers can process multiple high-resolution video streams with dual encoders and decoders supporting formats including H.265, H.264, VP9, and AV1. Capabilities include encoding up to 2× 4K60 or 16× 1080p30 and decoding up to 8K30 or 22× 1080p30. Camera support includes up to 16 lanes of MIPI CSI-2, enabling integration of multiple high-resolution sensors for advanced vision-driven AI systems.

Flexible I/O and Edge Deployment

When paired with carrier boards, the module supports PCIe Gen4, Gigabit Ethernet, USB 3.2, display outputs (DP 1.4, eDP, HDMI), and industrial interfaces such as CAN, SPI, I²C, and GPIO. This makes it adaptable across a wide range of autonomous machine use cases, from robotics and drones to industrial inspection and smart city applications.

Full NVIDIA AI Software Stack

Jetson AGX Orin runs the complete NVIDIA AI software stack, with support for frameworks like Isaac for robotics, DeepStream for vision AI, and Riva for conversational AI. Developers can accelerate workflows with Omniverse Replicator for synthetic data generation and the TAO Toolkit for fine-tuning pretrained models. A strong partner ecosystem provides additional software, carrier boards, sensors, and services to speed development and deployment.

• 275 TOPS AI performance for next-generation edge AI and robotics.
• NVIDIA Ampere GPU with 2048 CUDA cores and 64 Tensor Cores.
• 12-core Arm Cortex-A78AE CPU with 3MB L2 + 6MB L3 cache.
Dual NVDLA v2.0 and dual vision accelerators for AI inference and perception.
• 64GB LPDDR5 memory, 256-bit, 205 GB/s bandwidth.
• 64GB eMMC 5.1 onboard storage.
• Camera input: 16 lanes MIPI CSI-2 for multi-sensor support.
• Video encode: up to 2× 4K60 / 16× 1080p30 (H.265, H.264, AV1).
• Video decode: up to 8K30 / 22× 1080p30 (H.265, H.264, VP9, AV1).
• Display outputs: DP 1.4, eDP, HDMI 2.0.
• Connectivity via PCIe Gen4, USB 3.2, Gigabit Ethernet, plus CAN, I²C, SPI, GPIO.
• Configurable power range: 15W–60W.
• Compact module footprint: 100mm x 87mm with 699-pin connector.