AMD Versal™ AI Edge XQRVE2302 becomes the second radiation-tolerant device in the space-grade (XQR) Versal adaptive SoC portfolio to be qualified for spaceflight, having achieved Class B qualification. Derived from the US military specification MIL-PRF-38535, the completion of Class B qualification, together with the publishing of the production data sheet, allows customers to begin placing orders, with devices expected to begin shipping in the Fall.
AMD Versal AI Edge XQRVE2302 devices bring accelerated AI inferencing to space with integrated, enhanced AMD AI Engines (AIE) optimized for machine learning applications. Known as AIE-ML, these compute engines deliver enhanced support for data types prevalent in AI inferencing, offering 2X the INT8 and 16X the BFLOAT16 performance, with reduced latency compared to first-generation AI Engines. Additionally, local memory has doubled via new memory tiles providing high bandwidth shared memory access.
XQRVE2302 devices bring powerful computing into a small form factor (23mm x 23mm package). It is the industry’s first adaptive SoC for space applications to be offered in such a small form-factor package. The devices feature the same high-performance processor system as the larger Versal AI Core XQRVC1902 space-grade device, in less than 30% of the board area, resulting in a smaller footprint and a significant reduction in power consumption.
XQRVE2302 devices feature a dual-core Arm® Cortex®-A72 application processor and a dual-core Arm Cortex-R5F real-time processor, along with AIE-ML, DSP blocks, and FPGA programmable logic. With these features, developers can convert raw sensor data into useful information, making it ideal for edge processing in space, such as image detection and classification, autonomous navigation, and sensor data processing.
AI inferencing applications in space include telemetry anomaly detection, wildfire detection, vegetation and crop identification, and cloud detection, which is critical for earth imaging satellites to avoid wasting power on transmitting data when the subject is obscured by clouds.
Alpha Data, a leader in FPGA-based acceleration boards, recently announced a radiation-tolerant reference design for the AMD Versal AI Edge XQRVE2302 adaptive SoCs. Engineered for space-grade design, development, and deployment, the ADM-VB630 reference design enables fast, cost-effective development and rapid prototyping for space applications.
“The AMD Versal™ AI Edge XQRVE2302 is key in the ADM-VB630 delivering a powerful and efficient solution for next-generation satellite applications,” said Andrew McCormick, Technical Director and CTO, Alpha Data. “By enabling AI and advanced sensor processing in space, we are opening new frontiers for in-orbit intelligence and are bringing AI inferencing to space.”
AMD Versal XQR Series are designed for complementary roles, allowing them to co-exist in the same system design. AMD XQRVC1902 space-grade devices are designed to handle heavy-duty signal processing functions, while the smaller XQRVE2302 is built to tackle command and control functions, AI inferencing, and edge compute functions.
The Versal XQR Series also represents a major leap forward in enabling sophisticated on-board processing in space. Unlike other radiation-tolerant FPGAs, AMD Versal XQR adaptive SoCs support unlimited reprogramming during development and after deployment, including in-flight in the harsh radiation environment of space.
Developers can produce designs for XQR Versal devices using the AMD Vivado™ tool suite and Vitis AI software platform from a variety of developer tools and industry-standard frameworks, including RTL, C and C++, Matlab, Caffe, TensorFlow, PyTorch, and others.
AMD will showcase the device at the Alpha Data and Avnet Silica booths at the SpacE FPGA Users Workshop (SEFUW). Workshop attendees can watch AMD space architect Ken O’Neill’s keynote presentation on the AMD Versal AI Edge XQRVE2302 on Wednesday, March 26th at 9:10am CET.
Learn more about AMD Versal™ XQR adaptive SoCs for space.
# # #
