Aitech has developed the S-A1760 Venus, claimed to be the most powerful and smallest radiation-characterised space AI GPGPU. The new, small form factor system is rated for space flight and small satellite constellations used in NEO and LEO applications.
The COTS-based system is validated for usage in these short duration and NEO/LEO environments through the Series 300 level qualification standard that recognises the rad-tolerant requirements of space components and systems not utilised in deep space or long-haul applications.
Dan Mor, director, Video and GPGPU Product Line at Aitech, noted: “With the growing need for advanced imaging and data processing throughout space-rated applications, transitioning our powerful GPGPU-based AI supercomputers to this industry was a logical choice. By validating these space-rated, COTS-based systems with a clearly defined and recognised qualification level, we’re helping lead the charge in the development of commercial space applications and small sat cluster innovations.”
Based on the company's A176 Cyclone GPGPU system, it also uses the NVIDIA Jetson TX2i SoM that features the Pascal architecture with 256 CUDA cores and reaches 1TFLOPS of processing. This compact, rugged, space-rated system is excellent for high performance, small footprint environments. It offers the most advanced solution for onboard edge computing such as AI, deep learning, and video and signal processing for the next generation of short-duration spaceflight, NEO and LEO satellites applications.
The new AI supercomputer comprises a robust set of I/O interfaces, including Gigabit Ethernet, UART Serial, USB 2.0, CANbus and discretes, as well as DVI/HDMI output. Video capture incorporates an HD-SDI input with a dedicated H.264 encoder and eight RS-170A (NTSC)/PAL composite channels offered simultaneously.
The solution is a truly SFF SWaP optimised system. It meets MIL-STD-810H requirements for bench handling and conforms to VITA 47 for shock and vibration.