Impulse Embedded now offers the Aaeon Technology's BOXER-8640AI, its latest rugged, embedded box PC and the first to be powered by the NVIDIA flagship Jetson AGX Orin SoM, which provides a sixfold increase in AI performance over the last generation Jetson AGX Xavier.
With four PoE ports for connecting IP cameras, it is focused on Vision AI applications, including public safety, intelligent factories, and comes with extra I/O to allow integration with a broad variety of peripheral devices and sensors.
While the device's main focus is computer vision and video analytics, it provides a rich selection of I/O to support device control and monitoring tasks. This flexible new PC has 4x Gigabit PoE ports and 2x superspeed+ 10Gbps USB-C ports to connect high bandwidth peripherals while supporting legacy devices with a configurable RS-232/422/485 serial interface through a standard DB9 port. Also, mobile and robotics applications will welcome the extra DB9 CANbus port and the externally accessible expansion options for further GPIO.
The Jetson Orin module is supplied as standard with a sizeable 64GB of eMMC storage onboard, but for those requiring more space, the company has furnished the device with extra options, including a MicroSD card slot, an M.2 2280 M-key interface and a 2.5" SATA SSD/HDD bay, which makes this SoM an excellent option for those in poor connectivity areas.
The GPU is constructed around the current Ampere architecture seen in the NVIDIA RTX 3000 series of consumer-grade graphics cards, which provide a notable Performance Per Watt upgrade over the Volta architecture. The ampere GPU offers 1792 CUDA cores and 56 Tensor cores resulting in 200TOPS AI Compute performance. The SoM also provides an upgraded octa-core ARM Cortex-A78AE CPU and 32GB of onboard 256-bit LPDDR5 memory that offers an increase of 68GB/s bandwidth over the LPDDR4 found in Jetson AGX Xavier.
The PC supports the NVIDIA JetPack 5.0 SDK. This edition of the JetPack SDK incorporates the latest Jetson Linux Driver Package, Linux operating system, CUDA-X accelerated libraries and APIs for Deep Learning, Computer Vision, Accelerated Computing and Multimedia development. Moreover, this SDK also offers example code, documentation, and developer tools for the host computer and the developer kit, while supporting higher-level SDKs such as DeepStream for streaming video analytics, Isaac for robotics and Riva for conversational AI.