Xilinx ships its Versal AI Core, Versal Prime, key parts of its adaptive compute acceleration platform

Versal is Xilinx's effort to rewrite compute for artificial intelligence and machine learning.
Written by Larry Dignan, Contributor

Xilinx, which specializes in processors that can be programmed, said it has shipped its Versal AI Core and Versal Prime devices as it aims to create a new platform for adaptable computing.

Versal represents a first installment of its adaptive compute acceleration platform, or ACAP, that's designed to leapfrog CPUs, GPUs and FPGAs in the data center.

As detailed last month, Versal is Xilinx's effort to rewrite compute for artificial intelligence and machine learning. The general idea is that ACAP's hardware and software can dynamically change based on workloads for the data center, automobile, 5G, wired and defense industries.

Versal ACAP architecture includes a flexible network on a chip that integrates various parts of the platform and them programmable to developers, hardware engineers and data scientists.

Xilinx's strategy for Versal also revolves around software, code libraries, middleware and frameworks.

Victor Peng, CEO of Xilinx, said Versal ACAP 7nm process silicon was delivered by TSMC ahead of schedule and now resides with early access customers. Peng last month acknowledged that the Versal play may take time to develop, but so did GPUs.

Xilinx said Versal AI Core is designed to deliver better performance than CPU and GPU architectures. Versal AI Core is optimized for cloud, networking and autonomous technology such as vehicles. Versal Prime is a series that consists of hardware optimized to connectivity and in-line acceleration. 


 Key points about Versal AI Core and Versal Prime devices:

  • Each has dual-core Arm Cortex-A72 application processors and dual-core Arm Cortex-5F real-time processors.
  • 2 million logic cells of adaptable hardware.
  • 3,000 DSP engines optimized for high-precision floating point and low latencies.
  • Versa AI Core devices offer up to 400 AI engines designed for inference and advanced signal processing.

Xilinx said the Versal portfolio will have four additional series of devices designed for cloud, networking, wireless, edge computing and endpoint applications.

  • Enterprise AI in 2019: What you need to know
  • Survey: Tech leaders cautiously approach artificial intelligence and machine learning projects
  • Free PDF download: Managing AI and ML in the enterprise
  • Enterprise AI and machine learning: Comparing the companies and applications
  • The true costs and ROI of implementing AI in the enterprise
  • Machine learning and information architecture: Success factors
  • CIO Jury: 92 percent of tech leaders have no policy for ethically using AI                 
  • Editorial standards