Seminario: “Optimizing Data Communication on Large-Scale GPU Cluster” di Francesco De Simone

Abstract: "Recent years have witnessed an increasing growth in the usage of GPUs in large-scale compute clusters. GPU clusters can reach a high computational power, but their performance is limited by data communication: host CPU-to-GPU communication, as well as inter-node data movement. In this presentation, I will show how those bottlenecks are relevant in the Marconi100 cluster at CINECA, and how we can improve scalability by minimizing the data communication through better data partitioning."  

  Il collegamento alla ISISLab community della piattaforma Discord è disponibile qui o anche utilizzando il seguente QR Code:

Il seminario si terrà sul canale vocale #seminars della piattaforma Discord è disponibile qui o utilizzando il seguente QR Code: