Rugged, radiation-tolerant FPGAs bring AI capabilities to satellites and the battlefield.
Xilinx Inc. is expanding the market reach of its Versal family based on its adaptive compute acceleration platform, or ACAP, to include aerospace and military-grade devices that bring AI and machine learning capabilities to satellites and defense.
San Jose-based Xilinx has been steadily expanding its FPGA offerings, including last year’s release of a Versal ACAP series aimed at cloud and telecommunications network infrastructure deployments. Those devices offer a tripling of bandwidth and a doubling of computing density compared to its previous generation of FPGAs, primarily for accelerating data center workloads.
In conjunction with the chipmaker’s Adapt event, Xilinx this week released a portfolio of new Versal ACAP devices aimed at space and military applications. The AI core and edge devices for space dubbed XQR Versal would provide machine learning inference and on-board data processing to satellite and the growing number of commercial space applications.
The radiation-tolerant device follows up on the release last May of what Xilinx billed as the first 20-nm “space-grade” FPGA for satellite and other orbital applications. The XQRKU060 FPGA is also aimed at bringing machine learning capabilities to space payloads.
The new XQR ACAP is designed to support “full reconfiguration” of satellite systems while in orbit. The AI series devices use organic ball-grid array packaging technology and are qualified for space applications under Mil-Std-883 for radiation tolerance in low-Earth and geosynchronous orbits.
Xilinx said its space-grade XQR Versal AI core device is targeted for availability by the end of the second quarter 2022.
Meanwhile, the military-grade version of the Versal ACAP is a ruggedized platform described as a heterogenous multicore processor for providing AI and machine learning capabilities to military hardware. Along with handling battlefield conditions, Xilinx said the XQ can be reconfigured as it adapts to evolving algorithms. The military portfolio includes AI core, edge, “prime” and “premium” versions.
The military-grade family includes rugged packaging qualified to the Mil-Std-883 specs, and is rated at temperatures ranging from -55°C to +125°C.
Xilinx said initial release of the ruggedized Versal devices is scheduled for the first quarter of 2022.
This article was originally published on EE Times.
George Leopold has written about science and technology from Washington, D.C., since 1986. Besides EE Times, Leopold’s work has appeared in The New York Times, New Scientist, and other publications. He resides in Reston, Va.