Versal AI Edge ACAP Series to Deliver AI-Enabled Intelligence in Automotive, Robotics, Healthcare, and Aerospace Applications

Versal AI Edge series for enabling AI innovation

Xilinx has introduced the Versal AI Edge series for enabling AI innovation from the edge to the endpoint. The new device delivers 4X AI performances per watt than the GPUs and lower latency, delivering far more capable devices at the edge. It also delivers 10X greater compute density than the previous-generation adaptive SoCs enabling intelligent autonomous systems.

 

The new Versal AI Edge adaptive computer acceleration platforms (ACAPs) can serve various advanced applications including automated driving with the highest levels of functional safety, collaborative robotics, predictive factory and healthcare systems, and multi-mission payloads for the aerospace and defense markets.

 

The 4 X AI performances have been made possible as the portfolio has been designed with AI Engine-ML that delivers 4X machine learning compared to previous AI Engine architecture and integrates new accelerator RAM with enhanced memory hierarchy for evolving AI algorithms. The device comes with multiple safety standards required in various domains including industrial (IEC 61508), avionics (DO-254/178), and automotive (ISO 26262) markets.

 

The Versal AI Edge Series has been designed into the proven 7nm Versal architecture in miniature size for computing at low latency with 6W power efficiency. Connectivity blocks of the device range from LPDDR-4266, 32Gb/s transceivers to meet all the necessary protocols in edge applications, 40G multirate Ethernet, PCIe Gen4 with CCIX, and native MIPI support for vision sensors up to eight-megapixel resolution and beyond—critical to Level-2 ADAS and above. For delivering the flexibility to be used in a wide range of applications and workloads from edge to cloud, the new ACAPs allows changes at both hardware and software level.

 

Features of Versal AI Edge Series

  • Optimized AI Engines-ML deliver 4X performance/watt vs. GPUs
  • Native support for diverse ML data types: INT8, INT4, BFLOAT16
  • 4MB on-chip accelerator RAM extends memory hierarchy for AI performance
  • Programmable I/O to integrate any sensor, any interface
  • Adaptable Engines for sensor fusion and pre-processing
  • Intelligent Engines for AI, vision processing, and radar & LiDAR processing
  • Scalar Engines for embedded compute and real-time control
  • Architected to meet IEC 61508 and ISO 26262 safety standards

 

Note: More technical information can be found in the Versal AI Edge Series Product Brief linked at the bottom of this page.