Product of the Week: Advantech’s NVIDIA Jetson-based AIR-020 AI Inference System

March 15, 2022

Story

As the world’s demand for intelligence increases, real-time edge computing no longer refers simply to data processing. Increasingly, it implies AI.

In applications like traffic monitoring, defect/quality inspection, people counting, medical imaging, or autonomous mobile robots (AMRs), the demand for edge AI also means requirements for low latency, minimal power consumption, and sufficient thermal dissipation – all of course, at the highest performance levels possible.

The Advantech AIR-020 series edge AI boxes accomplish this by offering edge computing users a scalable hardware portfolio based on the NVIDIA Jetson hardware ecosystem. With SKUs based on the Jetson Nano, Jetson TX2 NX, and Jetson Xavier NX modules, the AIR-020 series delivers anywhere from 472 GFLOPs of inferencing performance at a 5W thermal design power (TDP) all the way to 21 TOPs at 15W in a 139 x 110 x 44.5 mm (5.47" x 4.33" x 1.75") package.

AIR-020 series SKUs include:

• AIR-020N – Based on the Jetson Nano, the AIR-020N equips a 128 CUDA core Maxwell  GPU and quad-core 1.43 GHz Arm Cortex-A57 CPU, 4 GB LPDDR4 memory, and 16 GB eMMC Flash for a benchmarked 512 GFOLPs of floating-point 16 performance.
• AIR-020T – Based on the Jetson TX2 NX, the AIR-020T includes 256 CUDA GPU cores, dual-core Arm Denver 2 and quad-core Arm Cortex-A57 cores running at 2 GHz alongside 4 GB LPDDR4 memory and 16 GB of eMMC Flash for a benchmark of 1.33 TFLOPS
• AIR-020X – Based on the Jetson Xavier NX, the AIR-020X leverages a 384 Volta CUDA GPU cores, 48 Tensor cores, a six-core Arm v8.2 Carmel CPU, 8 GB LPDDR4 memory, and 16 GB eMMC Flash for a benchmarked 21 TOPS of integer-8 resolution performance.

Across the series, users can expect I/O and interfaces that include one or more:

• USB: USB 3.2 Type A/C, MicroUSB
• Networking: GbE
• Display: HDMI
• Serial: RS-232/422/485, CANbus
• Expansion: Full-size MiniPCIe with Nano SIM Slot

These capabilities combine in vision-centric use cases like those mentioned above to enable performance rates of up to 1058 frames per second (fps).

The Advantech AIR-020 Edge AI System in Action

For AI-centric applications, however, up to 128 GB of external storage can be added in the included storage drive that plugs into an onboard M.2 slot, which allows larger AI models and data sets to be accessed by the processor at will.

On the software side, the platforms run a long-term support version of Ubuntu 18, which plays host to development using version 4.5.1 of the JetPack SDK and Advantech’s Edge AI Suite. The no-code, GUI-based Edge AI Suite includes an inferencing engine and support for the OpenVINO toolkit, pre-trained AI models, and a deployment wizard, while the JetPack SDK includes CUDA X-accelerated libraries and deep learning, computer vision, and multimedia APIs.

The platforms are also outfitted with security technologies like secure boot and version 2.0 of the Trusted Computing Group’s (TCG’s) Trusted Platform Module security element to protect sensitive information at rest on the device or in transit across the IoT infrastructure.

Together, the hardware and software capabilities facilitate some of the most advanced intelligence available for today’s intelligent far edge computing applications.

In operation, the AIR-020 series supports a 12-24V DC power input and can operate over a -10 to 55°C temperature range. It’s also vibration rated to 3 Grms @ 5 ~ 500 Hz.

Getting Started with the Advantech AIR-020 Edge AI System

The comprehensive software package delivered with Advantech’ AIR-020 series of edge AI inferencing systems also includes Linux board support packages (BSPs) for fast bringup so application development can begin right out of the box. Of course, it doesn’t stop there, with plenty of demos, tutorials, applications notes, and more available in both the Advantech Edge AI Suite and JetPack community ecosystems.

For more information or to get started, visit the AIR-020 product page at www.advantech.com/products/65f20c25-f6ef-4ab5-be3c-b7dfa7a833b3/air-020/mod_fcf216c8-3495-4809-b815-61dc008d53a4, request a quote here, or check out the resources below.

Additional Resources:

• Advantech AIR-020 Series Overview: https://www.advantech.com/products/65f20c25-f6ef-4ab5-be3c-b7dfa7a833b3/air-020/mod_fcf216c8-3495-4809-b815-61dc008d53a4
• Advantech AIR-020 Product Series Datasheets:
  • AIR-020N: https://advdownload.advantech.com/productfile/PIS/AIR-020/file/AIR-020N_DS(011022)20220118094020.pdf
  • AIR-020T: https://advdownload.advantech.com/productfile/PIS/AIR-020/file/AIR-020T_DS(011022)20220118094018.pdf
  • AIR-020X: https://advdownload.advantech.com/productfile/PIS/AIR-020/file/AIR-020X_DS(011022)20220118094018.pdf
• Press Release: https://www.advantech.com/resources/news/advantech-announces-air-020-ultra-compact-ai-inference-systems-with-nvidia-jetson-family