NVIDIA Jetson AGX Orin Module
NVIDIA Jetson AGX Orin™ Module — Up to 275 TOPS AI Performance — Ampere Architecture — 32GB & 64GB LPDDR5 The NVIDIA Jetson AGX Orin™ is the highest-performance module in...
Read MoreSKU:NVD-012
- Regular price
- Rs. 140,749.99
- Sale price
- Rs. 140,749.99
- Regular price
-
Express Shipping
Shipping Info
Fast delivery options for your convenience
Pick-Up in Store
Free
Select this option to pick up your order directly from our store. Fast and convenient!
NVIDIA Jetson AGX Orin™ Module — Up to 275 TOPS AI Performance — Ampere Architecture — 32GB & 64GB LPDDR5
The NVIDIA Jetson AGX Orin™ is the highest-performance module in the Jetson Orin family, delivering up to 275 TOPS of INT8 AI compute in a compact 100 × 87 mm form factor pin-compatible with Jetson AGX Xavier. Available in 32GB and 64GB LPDDR5 configurations, it is purpose-built for robotics, autonomous machines, and industrial edge AI at the highest performance tier.
Key Highlights
- Up to 275 TOPS AI Performance — Delivers more than 8× the AI compute of Jetson AGX Xavier, enabling the most demanding real-time inference workloads at the edge without any cloud dependency.
- NVIDIA Ampere Architecture GPU — Up to 2048 CUDA cores and 64 Tensor Cores at 1.3 GHz, enabling concurrent execution of deep learning, computer vision, and graphics pipelines in a single module.
- Dual NVDLA v2.0 Deep Learning Accelerators — Two dedicated hardware engines offload neural network inference from the GPU, increasing throughput and power efficiency for always-on inference tasks.
- High-Bandwidth LPDDR5 Memory — Up to 64GB of unified 256-bit LPDDR5 at 204.8 GB/s provides ample headroom for large model weights and high-resolution sensor data without memory bottlenecks.
- Multi-Camera & Multi-Sensor Input — Supports up to 6 simultaneous cameras (16 via virtual channels) over 16 MIPI CSI-2 lanes with D-PHY 2.1 / C-PHY 2.0, ideal for 360° perception systems.
- PCIe Gen4 High-Speed I/O — Up to 2×x8, 1×x4, and 2×x1 PCIe Gen4 lanes enable direct attachment of NVMe storage, FPGAs, accelerator cards, and high-speed radio modules.
- Hardware Video Engine — Dedicated encode and decode engines handle up to 2×4K60 encode and 3×4K60 decode (64GB) with H.265, H.264, AV1, and VP9, freeing the GPU entirely for inference.
- Full NVIDIA AI Software Stack — Ships with JetPack SDK and is compatible with Isaac, DeepStream, Riva, TAO Toolkit, and Omniverse Replicator — all optimised for Orin's hardware accelerators.
- Configurable Power Envelope — Scalable from 15W to 60W (64GB) lets you tune the performance-power trade-off for battery-operated or thermally constrained deployments.
- Drop-In Xavier Compatibility — The 699-pin Molex Mirror Mezz connector and 100 × 87 mm footprint are pin-compatible with Jetson AGX Xavier, simplifying hardware migration for existing product designs.
Technical Specifications
| Specification | Jetson AGX Orin™ 32GB | Jetson AGX Orin™ 64GB |
| AI Performance | 200 TOPS (INT8) | 275 TOPS (INT8) |
| GPU | 1792-core NVIDIA Ampere GPU with 56 Tensor Cores | 2048-core NVIDIA Ampere GPU with 64 Tensor Cores |
| GPU Max Frequency | 930 MHz | 1.3 GHz |
| CPU | 8-core Arm® Cortex®-A78AE v8.2 64-bit (2MB L2 + 4MB L3) | 12-core Arm® Cortex®-A78AE v8.2 64-bit (3MB L2 + 6MB L3) |
| CPU Max Frequency | 2.2 GHz | |
| Memory | 32GB 256-bit LPDDR5 — 204.8 GB/s | 64GB 256-bit LPDDR5 — 204.8 GB/s |
| Storage | 64GB eMMC 5.1 | |
| DL Accelerator | 2× NVDLA v2.0 | |
| DLA Max Frequency | 1.4 GHz | 1.6 GHz |
| Vision Accelerator | 1× PVA v2 | |
| CSI Camera | Up to 6 cameras (16 via virtual channels) — 16 MIPI CSI-2 lanes — D-PHY 2.1 / C-PHY 2.0 | |
| Video Encode | 1×4K60, 3×4K30, 6×1080p60, 12×1080p30 (H.265, H.264, AV1) | 2×4K60, 4×4K30, 8×1080p60, 16×1080p30 (H.265, H.264, AV1) |
| Video Decode | 1×8K30, 2×4K60, 4×4K30, 9×1080p60, 18×1080p30 (H.265, H.264, VP9, AV1) | 1×8K30, 3×4K60, 7×4K30, 11×1080p60, 22×1080p30 (H.265, H.264, VP9, AV1) |
| PCIe | Up to 2×x8, 1×x4, 2×x1 PCIe Gen4 (Root Port & Endpoint) | |
| USB | 3× USB 3.2 Gen2 (10 Gbps), 4× USB 2.0 | |
| Networking | 1× GbE, 1× 10GbE | |
| Display | 1× 8K60 multi-mode DP 1.4a / eDP 1.4a / HDMI 2.1 | |
| Other I/O | 4× UART, 3× SPI, 4× I2S, 8× I2C, 2× CAN, PWM, DMIC & DSPK, GPIOs | |
| Power | 15W – 40W | 15W – 60W |
| Form Factor | 100mm × 87mm — 699-pin Molex Mirror Mezz Connector | |
| Thermal | Integrated Thermal Transfer Plate | |
Which AGX Orin Module Is Right for You?
Both modules share the same PCB footprint, connector, and I/O configuration — the choice comes down to AI throughput, CPU core count, and memory capacity. The 32GB is well-suited for most single-domain robotics workloads, while the 64GB unlocks simultaneous multi-domain inference, larger transformer models, and higher-resolution multi-camera video pipelines.
| Criteria | Jetson AGX Orin™ 32GB | Jetson AGX Orin™ 64GB |
| AI Performance | 200 TOPS | 275 TOPS |
| GPU Cores / Frequency | 1792 cores @ 930 MHz | 2048 cores @ 1.3 GHz |
| CPU Cores | 8-core Arm A78AE | 12-core Arm A78AE |
| Memory | 32GB LPDDR5 | 64GB LPDDR5 |
| Max Power | 40W | 60W |
| Video Encode (max) | 1×4K60 / 12×1080p30 | 2×4K60 / 16×1080p30 |
| Best For | Single-domain robotics, AMRs, industrial inspection | Multi-domain AI, large model inference, autonomous vehicles, multi-camera analytics |
Common Applications & Use Cases
- Autonomous Mobile Robots (AMRs) — Simultaneous navigation, obstacle avoidance, and object detection pipelines run concurrently using the GPU and DLA engines, enabling real-time operation without cloud dependence.
- 3D Perception & LiDAR Fusion — High memory bandwidth and the PVA vision accelerator handle dense point clouds and multi-sensor fusion for depth estimation and scene reconstruction at robotics frame rates.
- Industrial Quality Inspection — Multi-camera input and NVDLA accelerators power high-throughput defect detection on production lines, replacing multiple dedicated vision processors with a single module.
- Multi-Channel Video Analytics — Hardware decoders handle up to 22 simultaneous 1080p30 streams (64GB), enabling large-scale intelligent camera networks with per-stream AI analytics via DeepStream.
- Natural Language Processing at the Edge — Sufficient compute and memory for running mid-size transformer models and large language model inference locally, enabling voice assistants and conversational AI without cloud latency.
- Autonomous Vehicles & Drones — Combines real-time video encode, multi-sensor fusion, and path planning in a power-constrained form factor for next-generation UAVs and ground vehicles.
- Medical & Surgical Robotics — Deterministic AI performance and broad I/O connectivity support image-guided intervention systems, endoscopy AI, and real-time surgical assistance applications.
- Smart Retail & Occupancy Analytics — DeepStream-based multi-camera pipelines perform crowd analytics, shelf monitoring, and checkout automation without streaming video off-device.
- Synthetic Data Generation — Omniverse Replicator integration lets teams generate labelled training datasets directly on Jetson hardware, reducing cloud data pipeline complexity for model development.
- Edge AI Inference Server — PCIe Gen4 and 10GbE allow the module to serve multiple upstream devices as a centralised on-premise inference endpoint, replacing rack-mounted GPU servers in constrained environments.
What's in the Box
- 1× NVIDIA Jetson AGX Orin™ Module (32GB or 64GB, as ordered)
- Integrated Thermal Transfer Plate (factory-attached to module)
Note: a carrier board is required to use this module and is not included. Accessories such as power supplies, cables, cases, and storage devices are sold separately and not included unless stated above.
Frequently Asked Questions
Is the Jetson AGX Orin module compatible with existing Jetson AGX Xavier carrier boards?
Yes. The AGX Orin uses the same 699-pin Molex Mirror Mezz connector and 100 × 87 mm footprint as the Jetson AGX Xavier, making it a drop-in module replacement on most existing AGX Xavier carrier boards. You should confirm with your carrier board manufacturer that power delivery and BSP support have been validated for AGX Orin, as some older designs may require a firmware update or minor hardware revision. NVIDIA's official Developer Kit carrier board is fully validated for both generations out of the box.
What power supply does the Jetson AGX Orin module require?
The AGX Orin module receives power through its carrier board rather than a direct onboard connector. The 32GB variant operates between 15W and 40W, while the 64GB variant can draw up to 60W. When using the official Jetson AGX Orin Developer Kit carrier board, a compatible DC power supply (typically 19V input, rated for your TDP tier) is required and sold separately. Always consult your carrier board documentation for the exact connector type, voltage, and current rating. Running the module at lower power modes (e.g. 15W) is useful for thermally constrained or battery-powered enclosures.
What operating system and software stack does the Jetson AGX Orin support?
The Jetson AGX Orin runs Ubuntu Linux via NVIDIA's JetPack SDK, which bundles the BSP, CUDA, cuDNN, TensorRT, and multimedia APIs into a single flashable image. JetPack 6.x (based on Ubuntu 22.04) is the current recommended release for AGX Orin. The full NVIDIA AI software stack — including Isaac SDK, DeepStream, Riva, TAO Toolkit, and Triton Inference Server — is available and optimised for Orin's hardware accelerators. Windows and macOS are not supported; the module is Linux-only.
What storage options are available, and can I add an NVMe SSD?
Both AGX Orin modules include 64GB eMMC 5.1 soldered on-module for the root filesystem. For expanded storage, the carrier board's PCIe Gen4 interface supports high-speed NVMe SSDs via M.2 Key M — available on the official Developer Kit carrier board. You can configure the system to boot directly from NVMe if your carrier board and BSP support it, which is strongly recommended for workloads involving high-throughput data logging or large datasets. The eMMC is sufficient for the OS and applications, but external NVMe significantly extends storage capacity and write endurance.
What accessories do I need to start using the Jetson AGX Orin module?
At minimum you need a compatible carrier board, a DC power supply rated for your TDP tier, and either a monitor or SSH access for initial setup. For the quickest start, NVIDIA's Jetson AGX Orin Developer Kit includes a full-featured carrier board with all standard ports exposed. Additional recommended accessories include a USB keyboard and mouse, a DisplayPort or HDMI monitor, an NVMe SSD for expanded storage, and a USB-A to Micro-B cable for flashing via NVIDIA SDK Manager. Camera modules, sensors, and enclosures are application-specific and sold separately.
How does the Jetson AGX Orin compare to the Jetson AGX Xavier?
The AGX Orin delivers up to 8× more AI performance than the Jetson AGX Xavier (275 TOPS vs approximately 32 TOPS), thanks to the NVIDIA Ampere GPU architecture, upgraded NVDLA v2.0 accelerators, a faster Arm Cortex-A78AE CPU cluster, and LPDDR5 memory with higher bandwidth. It adds AV1 codec support, C-PHY 2.0 CSI, and PCIe Gen4 — all absent on Xavier. Critically, it retains the same connector and form factor, making hardware migration straightforward for existing Xavier-based product designs.
How many GPIO, UART, I2C, SPI, and CAN interfaces does the module expose?
Via the 699-pin connector, the AGX Orin exposes 4× UART, 3× SPI, 4× I2S, 8× I2C, 2× CAN, PWM, DMIC, DSPK, and multiple GPIOs — the exact signals available at your system depend on how your carrier board routes them. Compared to the smaller Orin NX series, the AGX Orin provides substantially richer peripheral connectivity for complex embedded designs with multiple sensors and actuators. The 16-lane MIPI CSI-2 interface supports both D-PHY 2.1 and C-PHY 2.0 physical layers for maximum camera flexibility.
Is this module suitable for beginners, or is it aimed at experienced embedded developers?
The AGX Orin module is primarily aimed at professional and industrial developers building production-grade embedded AI systems. It does not include a carrier board, display, or power supply, requiring knowledge of embedded Linux, BSP flashing via NVIDIA SDK Manager, and hardware bring-up. Beginners are better served by the Jetson AGX Orin Developer Kit or the Jetson Orin Nano Super Developer Kit, which include carrier boards and are ready to use out of the box. The module format is intended for integration into custom hardware products.
What is the most common mistake when integrating the AGX Orin into a custom carrier board?
The most frequent issue is under-specifying the power delivery circuitry on the carrier board. The AGX Orin can draw up to 60W at peak, and instantaneous transient currents during GPU burst workloads can spike significantly higher — insufficient bulk capacitance or undersized power rails cause instability and unexpected reboots. A second common pitfall is neglecting to apply a proper thermal interface material (TIM) between the module's Thermal Transfer Plate and the system heatsink or chassis. NVIDIA's carrier board design guide and the AGX Orin module datasheet both provide detailed power sequencing and thermal guidelines that must be followed precisely during hardware design.
Where can I find official documentation, firmware updates, and community support for the Jetson AGX Orin?
All official documentation — including the module datasheet, design guide, JetPack release notes, and BSP source — is available at developer.nvidia.com/embedded/jetson-agx-orin. Firmware and JetPack SDK updates are distributed through NVIDIA SDK Manager and the NVIDIA L4T (Linux for Tegra) apt repository. The NVIDIA Developer Forums at forums.developer.nvidia.com host an active Jetson community for technical support, driver patches, and community-contributed projects. The jetson-inference GitHub repository maintained by Dusty Franklin provides practical inference containers and demos optimised for the Orin platform.
