HPC Tech Talk: Intel Xeon Phi x200 (Knights Landing) vs. NVidia Tesla P100 GPU and DGX-1 Deep Learning Appliances
| Wann |
14.07.2016 von 16:30 bis 17:45 |
|---|---|
| Wo | Hörssal Otto-Krayer-Haus / St.-Meier-Str./Albertstraße |
| Name | Dirk von Suchodoletz / eScience |
| Termin übernehmen |
vCal
iCal
|
Speaker: Hans Pabst, Application Engineer for High Performance and Throughput Computing
Title: Intel Xeon Phi x200 (Knights Landing): Performance Impressions and Best Practices
Subtitle: Case study focusing on Quantum Espresso and CP2K
Abstract: This presentation gives a brief overview and reference of Intel Xeon Phi x200 family (codenamed Knights Landing or KNL), and thereby focuses on ISA-level enhancements when compared to the previous generation of Intel’s Manycore Architecture. The talk gives special focus on Xeon Phi’s NUMA clustering and how to take advantage of the High Bandwidth Memory portion (with and without code changes). For the two application cases (physics domain), the presentation walks through five very basic steps of getting an application ready for Knights Landing: (1) building an optimized application, (2) selecting the cluster/memory mode, (3) process pinning, and thread affinitization, (4) performance profiling, and (5) code modernization and optimization.
Speaker: Ralph Hinsche (Nvidia), Business Development Manager
Title: Tesla P100 GPU and DGX-1 Deep Learning Appliances
Abstract:The binAC cluster in Tübingen, running on Tesla GPUs, is a step to accelerated computing and deep learning applications. This tech talk gives a short overview how the binAC cluster is built and continues with an outlook to the Tesla P100 platform and its performance features.