Over the years, AI systems have been at the front and center of many new technological advancements. They have played a crucial role in paving the way to solve a wide variety of tasks like motion planning, vehicle localization, speech recognition, automated parking, self-driving cars, and more. While early AI systems offered cloud-based services, these days – they have evolved to provide edge computing – thereby minimizing the amount of data transferred to the cloud, the number of gateways, bandwidth usage, and costs. It’s why edge AI systems, equipped with these high-performance computing capabilities are highly popular in embedded vision use cases.
As you may know, the NVIDIA® Jetson™ series of processors remain the most commonly-used host platform for embedded cameras – owing to their superior performance. And NVIDIA® Jetson AGX Orin™ is the latest release – with the Jetson AGX Orin™ 64GB and the Jetson AGX Orin™ 32GB modules.
Let us now deep dive into the performance of Jetson AGX Orin™ and look at its specifications and building blocks.
What is NVIDIA® Jetson AGX Orin™?
Built to accelerate next-generation edge AI applications, the Jetson AGX Orin™ 64GB module offers exceptional AI performance by delivering up to 275 TOPS with power configurable between 15W and 60W. On the other hand, the Jetson AGX Orin™ 32GB module delivers up to 200 TOPS with power configurable between 15W and 40W.
Figure 1: Jetson AGX Orin (Source: NVIDIA)
NVIDIA® Jetson™ AGX Orin – performance and technical brief
According to NVIDIA, the Jetson AGX Orin™ series offers up to eight times the performance of its predecessor Jetson AGX Xavier™. This series is also powered by the same AI software and cloud-native workflows used across other NVIDIA platforms.
The following figure shows the block diagram of the Orin System-on-Chip (SoC).
Figure 2: Block Diagram of Orin SoC (Source: NVIDIA)
Now, let us look at the technical specifications of Jetson AGX Orin™ 64GB and Jetson AGX Orin™ 32GB modules.
Technical specifications of Jetson AGX Orin series
The below table gives a detailed view of the technical specifications of NVIDIA Jetson AGX Orin:
|Jetson AGX Orin 32GB
|Jetson AGX Orin 64GB
|200 TOPS (INT8)
|275 TOPS (INT8)
|15W – 40W
|15W – 60W
|8-core Arm® Cortex®-A78AE v8.2 64-bit CPU 2MB L2 + 4MB L3
|12-core Arm® Cortex®-A78AE v8.2 64-bit CPU 3MB L2 + 6MB L3
|NVIDIA Ampere architecture with 1792 NVIDIA® CUDA® cores and 56 Tensor Cores
|NVIDIA Ampere architecture with 2048 NVIDIA® CUDA® cores and 64 Tensor Cores
|Max GPU Freq
|CPU Max Freq
|2x NVDLA v2.0
|DLA Max Frequency
|32GB 256-bit LPDDR5 204.8 GB/s
|64GB 256-bit LPDDR5 204.8 GB/s
|64GB eMMC 5.1
|Up to 6 cameras (16 via virtual channels*)
16 lanes MIPI CSI-2
D-PHY 2.1 (up to 40Gbps) | C-PHY 2.0 (up to 164Gbps)
|1x 4K60 | 3x 4K30 | 6x 1080p60 |
12x 1080p30 (H.265)
|2x 4K60 | 4x 4K30 | 8x 1080p60 |
16x 1080p30 (H.265)
|1x 8K30 | 2x 4K60 | 4x 4K30 | 9x
1080p60| 18x 1080p30 (H.265)
H.264, VP9, AV1
|1x 8K30 | 3x 4K60 | 7x 4K30 | 11x
1080p60| 22x 1080p30 (H.265)
H.264, VP9, AV1
|Up to 2 x8, 1 x4, 2 x1 (PCIe Gen4, Root Port & Endpoint)
3x USB 3.2
Single lane UFS
|1x 8K60 multi-mode DP 1.4a (+MST)/eDP 1.4a/HDMI 2.1
|4x USB 2.0
4x UART, 3x SPI, 4x I2S, 8x I2C, 2x CAN, DMIC & DSPK, GPIOs
|100mm x 87mm
699-pin Molex Mirror Mezz Connector
Integrated Thermal Transfer Plate
*Virtual Channel related camera information for Jetson AGX Orin™
is not final and subject to change by NVIDIA.
Major building blocks of Jetson AGX Orin™ series
GPU (Graphics Processing Unit)
The 64GB and 32GB Jetson AGX Orin™ SoC are based on the NVIDIA Ampere Graphic Processing Units (GPU) architecture. The Jetson AGX Orin™ 64GB module is composed of two Graphics Processing Clusters (GPCs), up to 16 Streaming Multiprocessors (SMs), and up to 8 Texture Processing Clusters (TPCs). The Jetson AGX Orin™ 32GB module contains 7 TPCs and 14 SMs.
As mentioned in the above table, the 64GB module has 2048 CUDA cores and 64 Tensor cores while the 32GB module has 1792 CUDA cores and 56 Tensor cores. This enables almost 3.7 times the performance of the GPU on Jetson AGX Xavier™.
DLA (Deep Learning Accelerator)
Deep Learning Accelerator is one of the major blocks to be considered for deep learning operations as this supports next-generation NVDLA 2.0, which is nine times the performance of NVDLA 1.0. It enables the GPU to run more complex networks and dynamic tasks.
CPU (Central Processing Unit)
Jetson AGX Orin™ series has Cortex-A78AE, which is a higher performance CPU with 8-core for the 32GB module and 12-core for the 64GB module. This CPU enables almost 1.9 times the performance of the 8-core NVIDIA Carmel CPU integrated on Jetson AGX Xavier.
e-con Systems has launched e-CAM82_CUOAGX – a 4K SONY STARVIS™ IMX485 ultra low light camera for NVIDIA® Jetson AGX Orin™
e-con Systems has been pioneering in the embedded camera space for 18+ years with a wide range of cameras supported on NVIDIA Jetson platforms (in addition to offering cameras compatible with other ARM-based platforms such as the NXP i.MX series, Raspberry Pi, Google Coral, etc. just to name a few). Being an elite partner to NVIDIA, we have always made sure to integrate all our cameras with the Jetson platform. The latest addition to this portfolio is e-CAM82_CUOAGX, a 4K Sony STARVIS IMX485 camera for NVIDIA Jetson AGX Orin.
This camera comes with a sensor size of 1/1.2” and a pixel size of 2.9µm making it a high sensitivity camera suitable for use in extreme low light conditions. With the ability to stream 4K@72 fps and 1080p@90fps, this is a high resolution and high frame rate camera that helps to capture better details of the scene. To learn more about the features of this AGX Orin camera and its applications, please visit the product page.
e-con Systems also extends its support for NVIDIA Jetson AGX Orin to all its camera modules with sensors from Sony, Onsemi, and Omnivision. Please visit the Camera Selector to look at our complete portfolio of products.
Hope this article gave you a fair bit of idea about what NVIDIA Jetson AGX Orin is. We have more coming up on AGX Orin. So stay tuned!
Meanwhile, if you have any questions on the topic, please feel free to leave a comment. And as always, if you are looking for help in integrating cameras into your products, please write to us at email@example.com.
Prabu is the Chief Technology Officer and Head of Camera Products at e-con Systems, and comes with a rich experience of more than 15 years in the embedded vision space. He brings to the table a deep knowledge in USB cameras, embedded vision cameras, vision algorithms and FPGAs. He has built 50+ camera solutions spanning various domains such as medical, industrial, agriculture, retail, biometrics, and more. He also comes with expertise in device driver development and BSP development. Currently, Prabu’s focus is to build smart camera solutions that power new age AI based applications.