Xilinx recently started shipping Versal series 7 nm multicore processors to select customers to address growing demands for Artificial Intelligence (AI) computing.
General availability will occur in the second half of 2019 for both the Versal AI Core and Versal Prime series, the company said in a release. Pricing was not announced.
Versal has developed the processors on its Adaptive Compute Acceleration Platform (ACAP), a heterogeneous compute platform that can be altered at the hardware and software levels to dynamically adapt to various applications and workloads. Possible customers include data centers, automotive developers, 5G wireless and wired networks and defense.
CEO Victor Peng said the Versal ACAP was produced ahead of schedule with 7nm technology by TSMC and represents “a major technology disruption.” Xilinx claims it is the first platform to combine software programmability with dynamically configurable domain-specific hardware acceleration. One key ingredient is intelligent engines for AI inference computing through up to 400 inference engines in the Versal AI Core series.
The Versal AI Core series has the ability to exceed AI inferencing capability by 8 times, in a comparison with the Nvidia V100.
Aside from Nvidia and Xilinx, AI inferencing is on the minds of all the major chip makers. “Nobody has captured the market in AI...it’s just really getting started,” said Nick Ni, director of product marketing for AI at Xilinx in an extensive interview with EETimes Europe.
He said that hardware has been the “bottleneck of mass deployment” and the chipmaker that gets the hardware right will win.