EN Bereichsnavigation EN

Intel Xeon Phi and NVIDIA Kepler Development Clusters

May 7th, 2013

Since April 2013 two test and development platforms are available at CSCS with Intel Xeon Phi and NVIDIA Kepler devices. Each compute node is composed of two Intel Sandy Bridge hosts and two accelerator devices.  The nodes are connected via InfiniBand FDR interconnect.  System details are available on our webpage user.cscs.ch and more specifically at

Dom (SB + K20) nodes:

http://user.cscs.ch/hardware/dom_gpu_cluster/index.html

Dommic (SB + KNS) nodes:

http://user.cscs.ch/hardware/dommic_intel_mic_cluster/index.html

Submitting batch jobs:

http://user.cscs.ch/running_batch_jobs/dom_gpu_dommic_mic/index.html

Please contact help@cscs.ch if you are interested in accessing this platform.

Day 4/3: Get “up to speed” with the Cray Cascade – videos online

May 3rd, 2013

Nathan Wikmann, Cray - Cray XC30: MPI Optimization Techniques

Alfio Lazzaro, Cray - Cray Scientific Libraries: Overview and Performance