Fanless GPU Computer Built On A Custom Jetson Carrier Board

When deploying AI frameworks at the edge, a rugged PC with GPU becomes indispensable. These ruggedised GPU computers are designed to withstand harsh environments, guaranteeing steady and dependable AI processing where traditional systems would quickly break down. Our fanless GPU computers built with NVIDIA Jetson have become increasingly popular over the years with advancements in computational performance being managed with techniques for passive cooling.

Designed with extended temperature components and featuring an optional heat spreader, the DSBOARD-ORNXS Jetson single-board computer (SBC) effectively enhances the computational capabilities of the NVIDIA Orin Nano and Orin NX modules for edge AI in harsh environments. Boasting a compact form factor and high-density Molex connectors for I/O breakout, the DSBOARD-ORNXS was selected for integration onboard a rugged GPU computer providing edge AI capabilities. Customised to the customer’s requirements, we included RS-232/422/485 serial control, 3G-SDI video capture, and GPIO control for putting the system to sleep or shutting down softly. Such customisation enables the small form factor NVIDIA Orin carrier board to efficiently handle tasks such as multi-sensor data processing, PTZ camera control, encoding/decoding, and AI inferencing, ensuring reliability in the customer’s challenging applications, which require a fanless GPU to eliminate all possible points of failure.

“We offer a wide range of NVIDIA Orin carrier boards suitable for various Edge AI applications, facilitating integration with a variety of sensors and cameras. With high-density I/O in a compact form factor, the DSBOARD-ORNXS is the perfect fit for defence and surveillance applications that demand advanced perception, object detection, and tracking capabilities. Built on a tiny footprint, we selected the DSBOARD-ORNXS for integration inside a ruggedised fanless GPU computer.” Andrew Whitehouse, Managing Director at Things Embedded.

TOPS, short for Tera Operations per Second, plays a pivotal role in assessing the processing capabilities of artificial intelligence systems. This metric measures the number of trillion operations an industrial edge AI system such as the Jetson Orin series can perform within a single second. TOPS holds particular importance when evaluating the performance of neural networks, which rely on intricate calculations. When you’re comparing our Edge AI computers, whether that’s a rugged computer with GPU or industrial NVIDIA Jetson PC, TOPS serves as a reliable indicator of the embedded system’s ability to swiftly handle complex tasks.

Measuring NVIDIA Jetson Orin vs its Xavier predecessor, you can see clearly that continuous advancements are being made to enhance the computational power in Edge AI applications. With an impressively compact design, these Orin Nano and Orin NX systems range from 20 to 100 TOPS in processing power, enabling them to handle a remarkable number of AI operations per second. According to NVIDIA, they are hailed as the world’s most potent AI computing solutions, making them a compelling choice for a wide array of processing tasks while maintaining minimal energy consumption. With power consumption as low as 7 watts, the Orin Nano emerged as the standout choice for our customers’ applications because of its remarkable energy efficiency. It proves to be an excellent fit for integration into their battery-powered rugged PC, where minimising energy usage was of paramount importance. Conversely, when a power source is readily available, the integration of the Orin NX takes the lead with its impressive performance capabilities. Integrating their fanless GPU computer with Orin NX allows them to efficiently handle demanding AI tasks with speed and precision.