NVIDIA Jetson AGX Orin: 275 TOPS, 2048 NVIDIA® CUDA® cores, 64 Tensor Cores, best AI performance of NVIDIA Jetson Family

NVIDIA Jetson AGX Orin: A Giant Leap Forward for Robotics and Edge AI with 275 TOPS, 2048 NVIDIA® CUDA® cores, 64 Tensor Cores

? Exciting news! NVIDIA Jetson AGX Orin Developer Kit is available to pre-order at Seeed now, be the first one get AGX Orin Dev kit now!

At GTC 2021 and 2022, NVIDIA introduced Jetson AGX Orin Module and Dev kit, the latest and the most powerful GPU so far in the Jetson family. Jetson AGX Orin is ready embedded into next-gen autonomous machines, robotics, or drones working for retail, manufacturing, transportation, AgTech, smart city, aerospace, etc. Let’s check out all the exciting features!

  • Most powerful member from Jetson family: NVIDIA Jetson AGX Orin is based on Ampere architecture with 2048 NVIDIA® CUDA® cores and 64 Tensor Core.
  • Compact leap forword to edge AI: same form factor at 100mm x 87mm and same pin compatibility as compared to Jetson AGX Xavier, delivering 8x the AI performance with 270 TOPS.
  • Support large and complex deep learning models to solve problems such as natural language understanding, 3D perception, and multi-sensor fusion.
  • Supported by NVIDIA JetPack™ and specific software platforms including Isaac for robotics, Clara for healthcare, and Metropolis for smart cities.

Stay tuned with us for more tutorials, applications and ecosystem updates around NVIDIA Jetson AGX Orin and all Jetson families!

Jetson AGX Orin?: NVIDIA

Jetson AGX Orin Module Technical Specifications

The Jetson AGX Orin is using the NVIDIA Ampere architecture GPU with 2048 NVIDIA® CUDA® cores and 64 Tensor cores and 12-core Arm Cortex-A78AE v8.2 64-bit CPU.

AGX Orin developer kit contains a Jetson AGX Orin module with heat sink and reference carrier board, 802.11ac/abgn wireless Network Interface Controller, USB-C power adapter, and cord together with a quick start and support guide.

AI Performance275 TOPS (INT8)
GPUNVIDIA Ampere architecture
with 2048 NVIDIA® CUDA® cores and 64 Tensor Cores
Max GPU Freq1 GHz
CPU12-core Arm® Cortex®-A78AE v8.2 64-bit CPU
3MB L2 + 6MB L3
CPU Max Freq2 GHz
DL Accelerator2x NVDLA v2.0
Vision AcceleratorPVA v2.0
Memory32GB 256-bit LPDDR5
204.8 GB/s
Storage64GB eMMC 5.1
CSI CameraUp to 6 cameras (16 via virtual channels*)
16 lanes MIPI CSI-2
D-PHY 1.2 (up to 40Gbps) | C-PHY 1.1 (up to 164Gbps)
Video Encode2x 4K60 | 4x 4K30 | 8x 1080p60 | 16x 1080p30 (H.265)
Video Decode1x 8K30 | 3x 4K60 | 6x 4K30 | 12x 1080p60| 24x 1080p30 (H.265)
UPHY2 x8 (or 1×8 + 2×4), 1 x4, 2 x1 (PCIe Gen4, Root Port & Endpoint)
3x USB 3.2
Single lane UFS
Networking1x GbE
4x 10GbE
Display1x 8K60 multi-mode DP 1.4a (+MST)/eDP 1.4a/HDMI 2.1
Other I/O4x USB 2.0
4x UART, 3x SPI, 4x I2S, 8x I2C, 2x CAN, DMIC & DSPK, GPIOs
Power15W | 30W | 50W
Mechanical100mm x 87mm
699-pin Molex Mirror Mezz Connector
Integrated Thermal Transfer Plate
Reference: NVIDIA, *Virtual Channel-related camera information for Jetson AGX Orin is not final and subject to change.

Jetson AGX Orin Reference Carrier Board

Camera16 lane MIPI CSI-2 connector
PCIex16 PCIe slot supporting x8 PCIe Gen4
RJ45Up to 10 GbE
M.2 Key Mx4 PCIe Gen 4
USB Type-C2x USB 3.2 Gen2 with USB-PD support
USB Type-A4x USB 3.2 Gen2
USB Micro-BUSB 2.0
DisplayPortDisplayPort 1.4a (+MST)
microSD slotUHS-1 cards up to SDR104 mode
Other40-pin header (I2C, GPIO, SPI, CAN, I2S, UART, DMIC)
12-pin automation header10-pin audio panel header
10-pin JTAG header
4-pin fan header
2-pin RTC battery backup connector
DC power jack
Force Recovery, and Reset buttons
Dimensions110mm x 110mm x 71.65mm
(Height includes feet, carrier board, module, and thermal solution)

We recommend you check the technical brief to learn more about the new Jetson AGX Orin module.

Get started with Jetson Software with necessary ML tools

Jetson Platform includes all the necessary ML tools to accelerate your next AI product quickly to market. At the base of the software stack, JetPack SDK includes the Board Support Package (BSP), with boot loader, Linux kernel, drivers, toolchains, and a reference file system based on Ubuntu.

Jetson Software comprises JetPack SDK, TAO & Pretrained AI Models from the NVIDIA NGC™ catalog, Triton Inference Server, NVIDIA Riva, DeepStream SDK, and NVIDIA Isaac. You could check the Jetson Software website for a more detailed explanation before getting your hands-on with Jetson AGX Orin.

Jetson software also supports containers and container orchestration, which brings cloud-native to the edge. NVIDIA JetPack includes NVIDIA Container Runtime with Docker integration, enabling GPU accelerated containerized applications on Jetson platform. With NVIDIA Triton™ Inference Server, you can also simplify models deployment for AI production at the scale.

Jetson Software Ecosystem ?: NVIDIA

Accelerate inferencing on the GPU using NVIDIA TensorRT and cuDNN

NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference. It includes a deep learning inference optimizer and runtime that delivers low latency and high throughput for deep learning inference applications.

With TensorRT, developers can focus on creating novel AI-powered applications rather than performance tuning for inference deployment.

? NVIDIA: TensorRT

cuDNN (CUDA Deep Neural Network library) is an acceleration library developed by NVIDIA for deep neural networks. It is a GPU acceleration library for deep neural networks. cuDNN enables high-performance GPU acceleration focused on training neural networks and developing software applications.

?NVIDIA: cuDNN Accelerated Frameworks

New Jetpack 5.0

According to the Jetson software roadmap, JetPack 5.0 (the current latest version is JetPack 4.6.1) will install Ubuntu 20.04, support Orin and Xavier Jetson modules. JetPack 5.0 preview release is targeted for Q1 2022 and the production release is targeted for the second half of 2022.

? NVIDIA: Compare Jetpack 4.6 and Jetpack 5.0

For next-gen autonomous machines

With Jetson AGX Orin Module that’s having 200 TOPS AI processing power (as compared to AGX Xavier which has 32 TOPS) on the developer kit, developers can deploy machine learning models in solving demanding challenges in edge AI applications for manufacturing, logistics, retail, service, agriculture, smart city, healthcare, and life science

Autonomous driving vehicles

Robotics such as in “real-time critical” autonomous driving vehicles in the enterprise setting which generally need to have a response within a millisecond and only possible with a lot of TOPS. There are six levels of self-driving cars – L0 (no automation), L1 (driver assistance), L2 (partial automation), L3 (conditional automation), L4 (high automation), and L5 (full automation) and the computing power required for each level of self-driving vehicle will be less than 10 TOPS for L2, 30 to 60 TOPS for L3, more than 100 TOPS for L4, and predictions of around 1,000 TOPS for L5.

Computational Requirements for different SAE Levels ?: GSAGLOBAL


Using Machine Learning capabilities on NVIDIA Jetson platform, it can improve the forecast accuracy for inventory management. This has a significant impact on optimizing the supply chain.  Machine Learning algorithms can be trained faster with NVIDIA’S RAPIDS open-source data processing and machine learning libraries, which enabled retailers to forecast the demand of the products for their stores more efficiently and restock accordingly according to the shopper trends. This helps in inventory cost savings at scale.

Retail Solution ?: NVIDIA


Manufacturing factories are making use of  logistics robots running on NVIDIA’s open Isaac robotics software platform architecture to automate the production capabilities. With the advanced logistics robots, this enabled the transport of  material autonomously, while manipulation robots select and organize parts.



The AI development pipeline can be accelerated with existing collected traffic data set with the pretrained TrafficCamNet model from NGC, and the NVIDIA TAO Toolkit on the Jetson NVIDIA Jetson edge AI platform. This has helped cut development time to just a few weeks, delivering a traffic safety solution to the city that automatically detects, counts, and measures speed and tracks pedestrian, bicycle, and vehicle occupancy.

TrafficCamNet ?: NVIDIA


The development and deployment of smart sensors with multimodal AI can be accelerated with the NVIDIA Clara™ Guardian, an application framework and partner ecosystem. Clara Guardian enables real-time insights to ensure safety through social distancing, fever screening, protective gear detection, and high-risk patient monitoring through fusing intelligent video analytics and conversational AI capabilities.

NVIDIA Clara Guardian ?: NVIDIA

At Seeed, you will find everything you want to work with NVIDIA Jetson Platform – official NVIDIA Jetson Dev Kits, Seeed-designed carrier boards, and kits, as well as third-party boards and accessories.

Seeed will continue working on the Jetson product line and will be ready to combine our partners’ unique skills with Seeed’s hardware expertise for an end-to-end solution.

About Author

5 thoughts on “NVIDIA Jetson AGX Orin: 275 TOPS, 2048 NVIDIA® CUDA® cores, 64 Tensor Cores, best AI performance of NVIDIA Jetson Family

  1. Is there a replacement for the Jetson Nano? For the purpose of education, budget matters.

Comments are closed.


March 2022