UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

cancel
Showing results for 
Search instead for 
Did you mean: 

Baidu Adopts Xilinx Kintex UltraScale FPGAs to Accelerate Machine Learning Applications in the Data Center

Xilinx Employee
Xilinx Employee
0 0 51.9K

 

Baidu SQL Accelerator.jpg Today, Xilinx announced that Baidu, China’s leading Internet search provider, is using Xilinx FPGAs to accelerate machine learning applications in their data centers located in China. Baidu’s Jian Ouyang discussed this work earlier this year at the Hot Chips conference held in Cupertino, California. (See “Baidu Takes FPGA Approach to Accelerating SQL at Scale” on the Nextplatform.com Web site. The Baidu Hot Chips paper is titled: “SDA: Software-Defined Accelerator for general-purpose big data analysis system.”) According to the Nextplatform.com article, “…Baidu sits on over an exabyte of data, processes around 100 petabytes per day, updates 10 billion Web pages daily, and handles over a petabyte of log updates every 24 hours.”

 

The Nextplatform.com article reports that Baidu developed its own FPGA board based on a Xilinx Kintex UltraScale KU115 FPGA paired with 8 to 32Gbytes of DDR4-2400 SDRAM. These boards automatically handle key SQL functions on demand. The article also contains two slides from Jian Ouyang’s Hot Chips presentation showing performance gains from the FPGA board. In one case, TPC-DS query3 runs 25x faster than the same function compiled from C++ and running in software. Terasort showed an 8x improvement. These are substantial performance gains and when you’re processing 10 billion Web pages a day, those sorts of numbers add up to big savings in data center capital expenses and energy costs.

 

Today’s press release also says that Baidu and Xilinx “are collaborating to further expand volume deployment of FPGA-based accelerated platforms.”