Xilinx claims world’s highest AI performance for Versal AI Edge chipset
Designed to enable AI innovation from the edge to the endpoint, the Versal AI Edge is Xilinx’s AI chipset which has four times the AI performance per watt than graphics processing units (GPUs) and 10x greater compute density than the Zynq Ultrascale+, says the company. The Versal AI Edge series is the world’s most scalable and adaptable portfolio for next-generation distributed intelligent systems, claims Xilinx.
The latest addition to Xilinx’s Versal ACAP portfolio of adaptive SoCs, the Versal AI Edge devices can be used for automated driving systems with the highest levels of functional safety (ASIL D). They can also be used for collaborative robotics, predictive factory and healthcare systems, and multi-mission aerospace and defence payloads, says Xilinx. The powerful AI engines have an enhanced memory hierarchy to deliver up to four times the machine learning (ML) compute performance per W compared with GPUs. They also have lower latency, continues Xilinx.
AI-enabled automated systems require high compute density that can accelerate whole applications from sensor to AI to real-time control. In addition to the increased compute density, enabling more intelligent autonomous systems, Versal AI Edge devices support multiple safety standards across automotive (ISO 26262), industrial (IEC 61508), and avionics (DO-254/178) markets.
The Versal AI Edge series takes the production-proven, 7nm Versal architecture and miniaturises it for AI compute at low latency, with power efficiency as low as 5W together with safety and security measures required in edge applications. The Versal AI Edge series matches the engine to the algorithm, with scalar engines for embedded compute, adaptable engines for sensor fusion and hardware adaptability, and intelligent engines for AI inference and advanced signal processing workloads ranging from beam forming and radar/lidar processing.
Connectivity blocks range from LPDDR-4266, 32Gbits per second transceivers to meet all the necessary protocols in edge applications, 40G multi-rate Ethernet, PCIe Gen4 with CCIX, and native MIPI support for vision sensors at up to 8Mpixel resolution, which is critical to Level-3 ADAS and above. The enhanced memory hierarchy which includes Accelerator RAM, makes the Versal AI Edge series suitable for a wider range of applications across numerous markets, according to Xilinx.
Versal AI Edge devices provide a design entry points for developers, including Vivado for hardware developers, the Vitis unified software platform for software developers, Vitis AI for data scientists, and complete software stacks for the platform’s target applications, including domain-specific operating systems, frameworks, and acceleration libraries.
Versal AI Edge series design documentation and support is available to early access customers, with shipments expected during the first half of 2022, and a roadmap for automotive- and defense-grade devices.