Comparing the performance of graph algorithms on AMD Xilinx FPGA and Nvidia GPUs

Summary

Graph processing has attracted tremendous interest for data analytics. However, energy consumption has become a limiting factor making energy-aware big data computing a necessity. Are you looking for a high performance and low power solution for graph analytics acceleration?

Vitis™ graph library offers the implementations of an extensive set of hardware kernels and software API functions of graph analytic algorithms to both hardware and software developers. It is open-source and accessible to all developers through GitHub and scalable across all AMD Xilinx platforms.

In this webinar, we will introduce the implementation of two key graph analytic algorithms: Breadth First Search (BFS) and Single Source Shortest Path (SSSP) on AMD Xilinx FPGAs and show that the solutions from Vitis graph library can deliver huge energy efficiency benefits, on AMD Xilinx Alveo™ data center accelerator cards vs. Nvidia Tesla T4 and P100 GPUs.

Key topics:

• Graph Analytics Challenges
• Performance benchmarks of BFS and SSSP algorithms between FPGA and GPU solutions

Whether you are a beginner doing graph analytics or a high-performance computing developer, the webinar will highlight the advantages of Vitis graph library for high performance and energy-efficient acceleration of big data graph analytics.

Featured Speakers

Cheng Ling

Product Marketing Manager, AMD