NVIDIA JETSON AGX ORIN DEVELOPER KIT(64GB)
NVIDIA Jetson AGX Orin Developer Kit — 275 TOPS Edge AI — 64 GB LPDDR5 — 2048-Core Ampere GPU The NVIDIA Jetson AGX Orin Developer Kit is NVIDIA's most powerful...
Read MoreSKU:NVD-015
- Regular price
- Rs. 225,299.99
- Sale price
- Rs. 225,299.99
- 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 Developer Kit — 275 TOPS Edge AI — 64 GB LPDDR5 — 2048-Core Ampere GPU
The NVIDIA Jetson AGX Orin Developer Kit is NVIDIA's most powerful edge AI platform, delivering up to 275 TOPS of AI performance — up to 8× more than Jetson AGX Xavier — from a compact module that shares the same form factor and software stack across the entire Orin family. Powered by a 12-core Arm Cortex-A78AE CPU, 2048-core Ampere GPU, dual NVDLA v2.0 deep-learning accelerators, and 64 GB of LPDDR5 unified memory, it is built for robotics, autonomous machines, industrial vision, and on-device generative AI inference backed by the full JetPack 6 SDK.
Key Highlights
- 275 TOPS AI Performance — Delivers up to 8× more INT8 throughput than the Jetson AGX Xavier, enabling real-time inference on large transformer, vision, and diffusion models at the edge without cloud dependency.
- 2048-Core NVIDIA Ampere GPU with 64 Tensor Cores — Full CUDA 12 support accelerates TensorRT deployment, mixed-precision workloads, and sparse tensor operations in a single unified memory space shared with the CPU.
- 64 GB 256-bit LPDDR5 — 204.8 GB/s Bandwidth — Shared GPU–CPU memory eliminates data-copy overhead, making it practical to run multi-billion-parameter models and large vision pipelines without batching constraints.
- Dual NVDLA v2.0 + PVA v2.0 Accelerators — Dedicated silicon for DNN inference and computer-vision pre/post-processing frees GPU cycles for the most latency-sensitive tasks in the pipeline.
- Multi-Camera & High-Throughput Video Pipeline — 16-lane MIPI CSI-2 input with hardware encode up to 4K60 and decode up to 8K30 supports multi-stream analytics and spatial AI systems with minimal CPU overhead.
- Configurable 15–60 W TDP with Four Power Modes — Software-selectable 15 W, 30 W, 50 W, and 60 W MAXN modes let you tune performance vs. power budget without changing hardware — ideal for both lab and deployed environments.
- PCIe Gen4, 10 GbE, Wi-Fi 5, Bluetooth 5.0 & Rich Expansion I/O — x16 PCIe Gen4 slot, M.2 Key M (x4 PCIe Gen4 NVMe), M.2 Key E, 10 Gigabit Ethernet, 6× USB 3.2 ports, DisplayPort 1.4a, and a 40-pin GPIO header cover virtually every sensor or peripheral integration scenario.
- Full NVIDIA JetPack 6 SDK Ecosystem — Ships ready to flash with JetPack 6.x (CUDA 12.6, TensorRT 10.3, cuDNN 9.3, VPI 3.2), Isaac ROS, TAO Toolkit, and DeepStream — the same APIs used in NVIDIA data-centre systems, so models move from cloud to edge without rewriting inference code.
Technical Specifications
| Specification | Details |
| AI Performance | Up to 275 TOPS |
| GPU | NVIDIA Ampere — 2048 CUDA cores, 64 Tensor Cores |
| CPU | 12-core Arm Cortex-A78AE v8.2 64-bit @ 2.2 GHz (3 MB L2 + 6 MB L3) |
| DL Accelerator | 2× NVDLA v2.0 |
| Vision Accelerator | PVA v2.0 |
| Memory | 64 GB 256-bit LPDDR5 — 204.8 GB/s bandwidth |
| Storage | 64 GB eMMC 5.1 + M.2 Key M (x4 PCIe Gen4) NVMe expansion |
| Video Encode | 2× 4K60 | 4× 4K30 | 8× 1080p60 | 16× 1080p30 (H.265) |
| Video Decode | 1× 8K30 | 3× 4K60 | 6× 4K30 | 12× 1080p60 | 24× 1080p30 (H.265) |
| Camera | 16-lane MIPI CSI-2 connector |
| Networking | 10 GbE (RJ45) + Wi-Fi 802.11ac + Bluetooth 5.0 (pre-installed M.2 module) |
| PCIe | x16 slot (x8 PCIe Gen4) + M.2 Key M (x4 PCIe Gen4) + M.2 Key E (x1 PCIe Gen4) |
| USB | 2× USB 3.2 Gen2 Type-C | 2× USB 3.2 Gen2 Type-A | 2× USB 3.2 Gen1 Type-A | 1× USB 2.0 Micro-B |
| Display | DisplayPort 1.4a with MST support |
| GPIO / Expansion | 40-pin header (I2C, GPIO, SPI, CAN, I2S, UART, DMIC), 12-pin automation header, 10-pin audio header, 10-pin JTAG, 4-pin fan, 2-pin RTC |
| Power | 15 W / 30 W / 50 W / 60 W MAXN (software-configurable), DC barrel jack 7–20 V |
| Dimensions | 110 mm × 110 mm × 71.65 mm |
Which Jetson Orin Is Right for You?
The Jetson Orin family scales from 67 TOPS up to 275 TOPS — all sharing the same Ampere GPU architecture, JetPack SDK, and software pipeline. Choose your module based on AI workload size, memory footprint, and power budget; the same Isaac ROS and TensorRT code runs on every variant without modification.
| Model | AI Performance | GPU Cores | CPU Cores | Memory | Power |
| Jetson AGX Orin Dev Kit | 275 TOPS | 2048 Ampere | 12-core A78AE | 64 GB LPDDR5 | 15–60 W |
| Jetson AGX Orin 32GB | 200 TOPS | 2048 Ampere | 8-core A78AE | 32 GB LPDDR5 | 15–40 W |
| Jetson Orin NX 16GB | 157 TOPS | 1792 Ampere | 8-core A78AE | 16 GB LPDDR5 | 10–40 W |
| Jetson Orin Nano 8GB | 67 TOPS | 1024 Ampere | 6-core A78AE | 8 GB LPDDR5 | 7–25 W |
Common Applications & Use Cases
- Autonomous Mobile Robots (AMR) — Run simultaneous SLAM, multi-sensor fusion, path planning, and obstacle avoidance in real time using all compute engines on a single board, replacing entire racks of compute in earlier robot architectures.
- Industrial Machine Vision & Quality Control — Process up to 16 simultaneous 1080p camera feeds with hardware decode and run defect-detection CNNs via NVDLA at throughputs that match high-speed production-line rates.
- Smart City & Traffic Analytics — Deploy multi-stream 4K video analytics for vehicle counting, licence-plate recognition, and anomaly detection at the edge without transmitting raw video to the cloud.
- Medical Imaging AI — Perform on-device inference for CT, MRI, and ultrasound analysis with the large unified memory pool enabling full-resolution 3D model inference that GPU-constrained devices cannot fit in memory.
- Autonomous Drones & UAVs — Execute low-latency real-time perception, obstacle avoidance, and landing-zone detection without cloud round-trips, with configurable TDP keeping airborne power budgets manageable.
- Generative AI & LLM Inference at the Edge — Run quantised large language models and diffusion models locally with TensorRT-LLM, enabling private, on-premises AI assistants and content generation without sending sensitive data off-device.
- Agricultural Robotics — Power multi-spectral crop-monitoring drones and harvesting robots that need vision, navigation, and actuation compute in a package capable of operating in field conditions.
- Retail & In-Store Analytics — Run anonymous people-counting, queue-length detection, and shelf-inventory AI simultaneously across multiple camera feeds without transmitting footage off-site.
- Robotics Research & Rapid Prototyping — The full set of carrier board interfaces — PCIe, GPIO, MIPI CSI, USB, CAN, I2S — lets research teams attach virtually any sensor or actuator and iterate without custom PCB work, with Isaac ROS providing ready-made perception primitives.
What's in the Box
- Jetson AGX Orin module (64 GB) with heatsink and fan on reference carrier board
- AzureWave AW-CB375NF Wi-Fi 802.11ac + Bluetooth 5.0 M.2 module (pre-installed)
- 90 W USB PD Type-C power supply with regional cords (US, EU, UK)
- USB Type-A to Type-C cable
- Quick Start Guide
Note: accessories such as NVMe SSDs, MIPI cameras, display adapters, and microSD cards are sold separately and not included unless stated above.
Frequently Asked Questions
What software and ML frameworks are compatible with the Jetson AGX Orin Developer Kit?
The Jetson AGX Orin runs the full NVIDIA JetPack 6 SDK stack, which includes CUDA 12.6, TensorRT 10.3, cuDNN 9.3, OpenCV, and VPI 3.2. It is fully compatible with Isaac ROS, DeepStream SDK, NVIDIA TAO Toolkit, and ROS 2 (Humble). Popular ML frameworks — PyTorch, TensorFlow, ONNX Runtime, and Triton Inference Server — all have validated ARM64 builds for Jetson. Because the Orin shares the same CUDA and TensorRT API versions as NVIDIA data-centre GPUs, models trained and optimised in the cloud can be deployed to the edge with minimal code changes.
What are the power requirements for the Jetson AGX Orin Developer Kit?
The developer kit is powered by the included 90 W USB Power Delivery Type-C adapter, which connects to the USB-C port on the carrier board. The module supports four software-selectable power modes: 15 W, 30 W, 50 W, and 60 W (MAXN), selectable via the nvpmodel utility or jtop. An alternative DC barrel jack (7–20 V input) is also present on the carrier board for integration into custom power systems. When building a production enclosure, account for peak wattage plus any peripherals connected to the PCIe slot or USB ports when sizing your power supply.
What operating system does the Jetson AGX Orin run?
The Jetson AGX Orin runs Jetson Linux (formerly L4T — Linux for Tegra), an Ubuntu-based distribution currently built on Ubuntu 22.04 LTS with Linux kernel 5.15. You flash it using NVIDIA SDK Manager from a host PC running Ubuntu 20.04 or 22.04 — the developer kit does not ship pre-flashed. JetPack 6.x is recommended for all new projects; JetPack 5.x (Ubuntu 20.04 base) remains available for projects requiring legacy compatibility. Red Hat Device Edge is also tested and supported as an alternative OS for enterprise deployments.
What storage options does the Jetson AGX Orin Developer Kit support?
The module includes 64 GB eMMC 5.1 on-module flash for the operating system and core applications. For expanded storage, the carrier board provides an M.2 Key M slot (x4 PCIe Gen4) supporting NVMe SSDs in the 2280 form factor — this is the recommended path for large datasets and model files, as Gen4 NVMe delivers far higher sequential throughput than the eMMC. A microSD slot (UHS-1, SDR104 mode) is also available for removable media. NVMe drives are not included in the box and are sold separately; a PCIe Gen4 NVMe 2280 SSD is strongly recommended for any serious inference or training workload.
What accessories do I need to get started with the Jetson AGX Orin Developer Kit?
Out of the box you need: a display connected via DisplayPort 1.4a (or a DP-to-HDMI adapter, sold separately), a USB keyboard and mouse, and a host PC running Ubuntu 20.04 or 22.04 for flashing JetPack via NVIDIA SDK Manager. The 90 W power supply and USB-C cable are included. For camera-based projects, USB cameras work immediately out of the box; MIPI CSI-2 cameras require an optional camera adapter board. An NVMe SSD (sold separately) is strongly recommended for working with large models or datasets beyond what the eMMC comfortably accommodates.
How does the Jetson AGX Orin compare to the Jetson AGX Xavier?
The Jetson AGX Orin delivers up to 8× more AI compute than the AGX Xavier — 275 TOPS vs. 30 TOPS. The GPU steps from 512-core Volta to 2048-core Ampere with Tensor Cores and sparse-tensor support not available on Xavier. The CPU moves from 8-core Carmel to a 12-core Cortex-A78AE at 2.2 GHz, approximately 1.7× faster per-thread. Memory doubles to 64 GB LPDDR5 with 1.4× higher bandwidth than Xavier's LPDDR4x. Both modules share the same 699-pin SOM connector, so many Xavier-compatible carrier boards are physically compatible with Orin — though software must be updated to JetPack 5 or 6.
How many GPIO pins and interfaces does the 40-pin header provide?
The standard 40-pin expansion header (2×20, 2.54 mm pitch, Raspberry Pi HAT-compatible pinout) provides I2C (×2), SPI (×2), UART (×2), I2S, CAN, GPIO, PWM, 3.3 V, and 5 V supply rails — exact pin allocation depends on the device-tree overlay loaded at boot. Beyond the 40-pin header, the carrier board adds a 12-pin automation header (SPI, DMIC, GPIO), a 10-pin audio header, a 10-pin JTAG debug header, a 4-pin fan header, and a 2-pin RTC battery connector. This combination lets you attach motor drivers, IMUs, CAN bus networks, and audio peripherals simultaneously without an I/O hub.
Is the Jetson AGX Orin Developer Kit suitable for beginners or only advanced users?
The developer kit targets professional developers, researchers, and advanced students — it is not a general-purpose single-board computer. Familiarity with Linux CLI, Python or C++, and a working knowledge of ML frameworks is expected. That said, NVIDIA provides free Jetson AI Courses on the NVIDIA Deep Learning Institute (DLI) covering JetPack setup, TensorRT optimisation, and Isaac ROS basics. The official Hello AI World tutorial gets a real-time object-detection pipeline running in under an hour. Developers comfortable with Raspberry Pi who want to move into serious AI inference will find the learning curve worthwhile but should budget time for the initial JetPack flashing and SDK environment setup.
What is the most common mistake when setting up the Jetson AGX Orin for the first time?
The most frequent pitfall is attempting to flash JetPack from a Windows or macOS host — NVIDIA SDK Manager only runs on x86_64 Ubuntu 20.04 or 22.04. A second common issue is forgetting to enter Force Recovery mode before connecting to the host PC — hold the Force Recovery button while pressing Power; the board will not appear as a USB device otherwise. Finally, many users do not expand the root partition after flashing and run out of space when installing JetPack components on the 64 GB eMMC — installing immediately to an NVMe SSD or running resize2fs on the eMMC partition right after first boot eliminates this issue entirely.
Where can I find documentation, community support, and firmware updates for the Jetson AGX Orin?
Official documentation lives on the NVIDIA Developer portal (developer.nvidia.com/embedded), including the Jetson AGX Orin User Guide, JetPack release notes, and the Jetson Linux driver package. Firmware and JetPack updates are distributed via NVIDIA SDK Manager (GUI installer) or the Jetson Linux archive for command-line flashing. Community support is active on the NVIDIA Developer Forums under the Jetson & Embedded Systems category — most hardware and software questions have existing threads. The jetson-containers project on GitHub provides a curated library of ready-to-run Docker containers for LLMs, diffusion models, vision pipelines, and robotics workloads, saving significant environment-setup time.
