NVIDIA® Jetson Orin™: A Breakthrough in Industrial Edge AI

By Louisa Rochford News Published 6 December 2023
Industry Insight Industrial Computing Embedded Systems Product Highlight AI

First released in August 2022, NVIDIA’s latest and greatest addition to the Jetson series marks a staggering leap forward in AI capabilities. The Jetson Orin, available in three distinct versions—that being Nano, NX and AGX—is a powerful series of System-on-Modules that are revolutionising Edge AI as we know it.

These powerful modules, ranging from 20 to 275 TOPS with an impressively small form factor, can manage a remarkable number of AI operations per second. This makes them, in NVIDIA’s words, the world’s most powerful AI computers and a formidable option for a vast range of processing tasks with minimal energy consumption.

 

The Orin Timeline

The timeline for the Jetson Orin dates back to March 2018, when the next-gen SoC (System-on-Chip) ‘Orin’ was first announced as being in development as the successor to the Jetson Xavier. The first device to be released in the Orin series was the Jetson AGX Orin Developer Kit, released in March 2022. That same year, the first standalone module of the series, the AGX Orin 32GB, was released in August 2022.

Next was the Orin NX 16GB module, released in late January 2023, followed not long after by the Orin Nano. Orin Nano was announced in September 2022 at the GPU Technology Conference 2022 and released in March 2023 in both 4GB and 8GB variants. In the same month, the 64GB variant of the AGX Orin and the 8GB variant of the Orin NX were also released.

Finally, the AGX Orin Industrial module was announced at COMPUTEX 2023 and became available in May 2023.

Jetson Orin Road Map

Orin Model Specs

The specs for each Orin module are outlined in the table below:

Feature Jetson Orin Nano 4GB Jetson Orin Nano 8GB Jetson Orin NX 8GB Jetson Orin NX 16GB Jetson AGX Orin 32GB Jetson AGX Orin 64GB Jetson AGX Orin Industrial
AI Performance 20 TOPS 40 TOPS 70 TOPS 100 TOPS 200 TOPS 275 TOPS 248 TOPS
GPU 512-core NVIDIA Ampere architecture GPU with 16 Tensor Cores 1024-core NVIDIA Ampere architecture GPU with 32 Tensor Cores 1024-core NVIDIA Ampere architecture GPU with 32 Tensor Cores 1024-core NVIDIA Ampere architecture GPU with 32 Tensor Cores 1792-core NVIDIA Ampere architecture GPU with 56 Tensor Cores 2048-core NVIDIA Ampere architecture GPU with 64 Tensor Cores 2048-core NVIDIA Ampere architecture GPU with 64 Tensor Cores
GPU Max Freq. 625 MHz 625 MHz 765 MHz 918 MHz 930 MHz 1.3 GHz 1.2 GHz
CPU 6-core Arm® Cortex®-A78AE v8.2 64-bit CPU 6-core Arm® Cortex®-A78AE v8.2 64-bit CPU 6-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU 8-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU 8-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU 12-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU 12-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU
CPU Max Freq. 1.5 GHz 1.5 GHz 2 GHz 2 GHz 2.2 GHz 2.2 GHz 2.0 GHz
DL Accelerator - - 1x NVDLA v2.0 2x NVDLA v2.0 2x NVDLA v2.0 2x NVDLA v2.0 2x NVDLA v2.0
DL Max Freq. - - 614 MHz 614 MHz 1.4 GHz 1.6 GHz 1.4 GHz
Memory 4GB 64-bit LPDDR5 34GB/s 8GB 128-bit LPDDR5 68GB/s 8GB 128-bit LPDDR5 102.4GB/s 16GB 128-bit LPDDR5 102.4GB/s 32GB 256-bit LPDDR5 204.8GB/s 64GB 256-bit LPDDR5 204.8GB/s 64GB 256-bit LPDDR5
(+ ECC) 204.8GB/s
Storage -
Supports external NVMe
-
Supports external NVMe
-
Supports external NVMe
-
Supports external NVMe
64GB eMMC 5.1 64GB eMMC 5.1 64GB eMMC 5.1
Power 7w – 10w 7w – 15w 10w – 20w 10w – 25w 15w – 40w 15w – 60w 15w – 75w

 

Powerful AI Operations

TOPS, or Tera Operations per Second, serves as a pivotal metric in assessing the processing power of artificial intelligence and digital signal tasks. This metric quantifies the number of trillion operations a system can execute within a second, offering a measure of the processing speed and efficiency for different devices.

TOPS is especially crucial in evaluating the performance of neural networks, where intricate calculations are vital. In the context of high-performance computing, such as edge devices or dedicated AI processors, TOPS clearly indicates a system’s capacity to handle complex tasks swiftly.

The relevance of TOPS as a benchmark in today’s industrial activities highlights its significance in shaping the landscape of modern computing, and the Orin serves as proof that advancements are always being made to improve the computational power of our modern technology.

 

Jetson Xavier—How Does It Compare?

The Orin series of computers succeeds the Xavier series with a substantial improvement in its specs. For example, comparing the Jetson Xavier NX 16GB to the Orin NX 16GB, the Orin can achieve an astonishing 79 trillion more operations per second with the same amount of memory. Additionally, power consumption has been optimised, as evidenced by the 20W max consumption from the Xavier model, compared to the 25W max consumption from the Orin despite achieving five times the TOPS.

The AGX Orin 64GB model, the most powerful from NVIDIA’s modules, improves substantially over the Xavier, with 275 TOPS—eight times more powerful than the AGX Xavier 64GB model.

For a detailed comparison of the Xavier specs compared to the Orin, see the table below:

Feature Jetson Xavier NX 16G Jetson Orin NX 16GB Jetson AGX Xavier 32GB Jetson AGX Orin 32GB Jetson AGX Xavier 64GB Jetson AGX Orin 64GB
AI Performance 21 TOPS 100 TOPS 32 TOPS 200 TOPS 32 TOPS 275 TOPS
GPU 384-core NVIDIA Volta™ GPU with 48 Tensor Cores 1024-core NVIDIA Ampere GPU with 32 Tensor Cores 512-core NVIDIA Volta™ GPU with 64 Tensor Cores 1792-core NVIDIA Ampere GPU with 56 Tensor Cores 512-core NVIDIA Volta™ GPU with 64 Tensor Cores 2048-core NVIDIA Ampere GPU with 64 Tensor Cores
GPU Max Freq. 1100 MHz 918 MHz 1377 MHz 930 MHz 1377 MHz 1.3 GHz
CPU 6-core NVIDIA Carmel Arm® v8.2 64-bit CPU 8-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU 8-core NVIDIA Carmel Arm® v8.2 64-bit CPU 8-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU 8-core NVIDIA Carmel Arm® v8.2 64-bit CPU 12-core NVIDIA Arm® Cortex A78AE v8.2 64-bit CPU
CPU Max Freq. 1.9 GHz 2 GHz 2.2 GHz 2.2 GHz 2.2 GHz 2.2 GHz
DL Accelerator 2x NVDLA v1.0 2x NVDLA v2.0 2x NVDLA v1.0 2x NVDLA v2.0 2x NVDLA v1.0 2x NVDLA v2.0
DL Max Freq. 1.1 GHz 614 MHz 1.4 GHz 1.4 GHz 1.4 GHz 1.6 GHz
Memory 16GB 128-bit LPDDR4x 16GB 128-bit LPDDR5 102.4GB/s 32GB 256-bit LPDDR4x 32GB 256-bit LPDDR5 204.8GB/s 64GB 256-bit LPDDR4x 64GB 256-bit LPDDR5 204.8GB/s
Storage 16GB eMMC 5.1 -
Supports external NVMe
32GB eMMC 5.1 64GB eMMC 5.1 32GB eMMC 5.1 64GB eMMC 5.1
Power 10w - 20w 10w - 25w 10w - 30w 15w - 40w 10w - 30w 15w - 60w

 

Orin for Industrial AI Purposes

The AGX Orin Industrial module is especially suited for settings with harsh environmental factors, with the computational power required for advanced artificial intelligence tasks. This model comes with the same improved TOPS to handle trillions of AI operations per second as the other Orin models, along with an extended temperature range, enhanced shock and vibration resistance, and support for Error Correction Code (ECC) memory. This makes it durable for environments where common areas of failure, such as those caused by extreme heat, dust, and shock, can be addressed.

Edge AI computers, such as Advantech’s MIC-711-OX which features NVIDIA’s Jetson Orin NX 16GB module, can be used for a plethora of uses in industries that are innovating and adapting every day. This includes manufacturing, where AI can be used to detect defects during quality inspections, or in driverless vehicles and AGVs (Automated Guided Vehicles).

Additionally, industrial AI is used in the railway industry to offer innovative solutions that enhance performance in areas such as efficiency, safety, and daily operations. Predictive maintenance can be optimised using AI, which analyses historical data to predict when maintenance on the rails and overhead cables is likely to be required, making it a proactive process. AI can also vastly improve safety by identifying when trespassers and foreign objects are on the tracks or anticipate a landslip due to changes in the wayside, which can then prompt dynamic action to prevent incidents. Additionally, it is currently being used to track ticket evaders on the railway and to detect overcrowding inside train carriages, allowing train operating companies (TOCs) to estimate passenger numbers more accurately.

 

AI Inference with Industry Insight

As a provider of powerful and innovative solutions such as NVIDIA’s Orin series, we at Impulse empower the industrial sector in the UK with the latest artificial intelligence processing capabilities. With a proven track record in the industry and a dedication to quality, we bring a wealth of experience in providing robust embedded computing solutions that are tailored to diverse industrial needs.

Whether deployed in manufacturing, automation, rail or other industries, our devices represent the technological advancement that allows businesses to thrive in an era where AI is becoming increasingly prevalent. Our responsive customer support ensures a smooth and efficient experience from device selection to post-sales assistance and beyond.

For more information regarding our Edge AI computers, contact us today.