The AMD Versal™ AI Edge XQRVE2302 has officially achieved Class B qualification, making it the second radiation-tolerant device in the space-grade (XQR) Versal adaptive SoC lineup to be cleared for spaceflight. Derived from the U.S. military standard MIL-PRF-38535, this qualification—along with the release of the production data sheet—enables customers to place orders, with shipments expected to begin in the fall.
Designed for space-based AI inferencing, the XQRVE2302 features advanced AMD AI Engines (AIE-ML), optimized for machine learning tasks. These processing units provide double the INT8 and 16 times the BFLOAT16 performance of previous AI Engines, while reducing latency. Enhanced local memory, supported by new high-bandwidth memory tiles, further improves data processing capabilities.
This compact device, housed in a 23mm x 23mm package, is the first adaptive SoC designed for space applications in such a small form factor. Despite its reduced footprint—occupying less than 30% of the board area of its larger counterpart, the Versal AI Core XQRVC1902—it retains a powerful processing system while significantly lowering power consumption.
Equipped with 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, the XQRVE2302 enables on-board edge processing in space. It is ideal for applications such as real-time image recognition, autonomous navigation, and sensor data analysis. AI-driven tasks in space include anomaly detection in telemetry, wildfire monitoring, vegetation and crop classification, and cloud detection—critical for optimizing earth observation satellites by preventing unnecessary data transmission when cloud cover obstructs the view.
To accelerate development, Alpha Data, a leader in FPGA-based acceleration solutions, has introduced a radiation-tolerant reference design for the XQRVE2302. The ADM-VB630 reference board provides a cost-effective platform for designing and prototyping space applications.
“The AMD Versal™ AI Edge XQRVE2302 is central to the ADM-VB630, delivering a powerful and efficient solution for next-generation satellite missions,” said Andrew McCormick, Technical Director and CTO at Alpha Data. “By enabling AI-driven sensor processing in space, we are advancing in-orbit intelligence and bringing AI inferencing capabilities to space environments.”
The Versal XQR Series is designed for complementary roles, allowing different devices to work together within the same system. The larger XQRVC1902 handles complex signal processing, while the compact XQRVE2302 is optimized for command and control, AI inferencing, and edge computing tasks.
Unlike traditional radiation-tolerant FPGAs, AMD Versal XQR adaptive SoCs allow unlimited reprogramming both during development and after deployment—even in orbit, where they must withstand the challenges of space radiation. Developers can create designs for XQR Versal devices using the AMD Vivado™ tool suite and Vitis AI software platform. These tools support a range of programming languages and frameworks, including RTL, C, C++, Matlab, Caffe, TensorFlow, and PyTorch.
AMD will showcase the XQRVE2302 at the SpacE FPGA Users Workshop (SEFUW) at the Alpha Data and Avnet Silica booths. Attendees can also hear AMD space architect Ken O’Neill discuss the new device in a keynote presentation on Wednesday, March 26, at 9:10 AM CET.
