
By deploying the FPGAs, SKT gained up to five times higher performance in the ASR application compared to GPUs, and 16 times better performance-per-watt.
The FPGA-based accelerators use Xilinx FPGA add-in cards for existing CPU-only servers. The ASR servers then accelerate multiple voice service channels. One FPGA card provides more than five times the performance of a single server.
SKT plans to demonstrate the FPGA-based AI solution at the Xilinx Developer Forum (XDF) in Silicon Valley on October 1-2, and in Beijing on October 16.
More information
Related news
Xilinx Zynq UltraScale+ SoC module with two memory channels
Xilinx buys Chinese machine learning startup
Xilinx Zynq UltraScale+ MPSoC module
X-Ware IoT Platform to support Xilinx’ Zynq UltraScale+ MPSoCs
