FPGAs achieve 5X/16X performance-per-watt over GPUs

This is the first commercial adoption of FPGA accelerators in the AI domain for large-scale data centres in South Korea.

The FPGA-based accelerator lowers the total cost of ownership (TCO) of ASR application servers by populating existing CPU-only servers with efficient Xilinx FPGA add-in cards. The ASR servers easily and simply accelerate multiple voice service channels with Xilinx FPGA cards in their empty slots.

One FPGA card provides more than five times the performance of a single server, resulting in substantial cost savings. This performance and TCO advantage benefits SK Telecom and its customers.

“Over many years we have seen the shape of the industry evolve, and we are proud to be at the forefront of developing AI accelerators. By designing our solution based on the Xilinx KCU1500 board and our own bitstream image, we have developed a cost-effective, high-performance application,” said Kang-Won Lee, senior vice president, software research and development centre at SKT.

The adaptive nature of Xilinx FPGAs enables fast deployment of custom hardware accelerators for the rapidly evolving field of AI and deep learning.

Moreover, FPGAs provide higher performance and lower latency at lower power when compared to CPUs and GPUs. SK Telecom joins a fast-growing list of high-profile commercial data centres that have deployed FPGAs for compute acceleration.

“We are delighted to have the opportunity to deploy Xilinx FPGAs to SKT’s AI data centre, a first in South Korea,” said Manish Muthal, vice president, marketing for data centre at Xilinx.

“Xilinx Kintex UltraScale KCU1500 FPGAs show that Xilinx has a major competitive edge in applications. Xilinx will continue to focus its technological capabilities and innovations on data centre acceleration.”

SKT plans to demonstrate its FPGA-based AI solution at the Xilinx Developer Forum (XDF) being held in Silicon Valley on 1st to 2nd October, and in Beijing on the 16th October.