Baidu takes FPGA approach to boost machine learning apps

Article By : Xilinx

Xilinx FPGAs deliver a 10-20X performance/watt improvement.

Chinese search giant Biadu is deploying Xilinx field programmable gate arrays (FPGAs) to speed machine learning applications in their data centres in China.

As rapid growth of emerging applications begins to drive up computational workloads, data centres are turning to application accelerators to keep up with the demands for greater throughput at low latency while retaining practical power levels. The two companies are collaborating to further expand volume deployment of FPGA-based accelerated platforms.

Xilinx FPGAs deliver the power efficiency that makes accelerators practical to deploy throughout the data centre and can deliver a 10-20X performance/watt improvement. Baidu-optimised FPGA platforms are tuned for machine learning applications such as image and speech recognition.

"Acceleration is essential to keep up with the rapidly increasing data centre workloads that support our growth," said Baidu executive director Yang Liu.

The platforms will also be leveraged in Baidu's initiative to develop commercially viable autonomous cars.