CSC will use the system in collaboration with Finnish researchers and the PRACE community to leverage the latest processing technologies from Intel and Nvidia. These new technologies promise significant improvements in cost-effectiveness and energy-efficiency compared to traditional processors.
The system will be deployed in two stages at CSC's data centre in Kajaani, Finland. The first stage is expected to be available for pilot users by the end of the year and it comprises 44 nodes with 88 Intel Xeon Phi 7120X coprocessors. Performance will be nearly doubled in early 2014 with the addition of a second stage using the latest Nvidia Tesla K40 GPUs. This system will be the very first in the world to integrate a combination of Intel Xeon Phi coprocessors and Nvidia Tesla K40 GPUs, and is intended to serve as a demonstrator for the use of both technologies.
When the demonstrator is complete, its peak performance will be on a par with CSC's Sisu - in its current configuration before the 2nd phase upgrade, the most powerful supercomputer in Finland, and among the most powerful in northern Europe.
It is also expected that the system will be one of the most energy-efficient supercomputer systems in the world. This is made possible by combining the most efficient processing technologies, cooling solutions and data centre infrastructure available today:
"Bull is strengthening CSC's supercomputing service portfolio as one of the leading providers of accelerator-based processing capabilities for pilot users in Finland and across Europe", stated Kimmo Koski, Managing Director of CSC.
"Bull is very proud to provide CSC with this pioneering system that combines the two leading accelerator technologies in the same hot water cooled environment. We are very happy that yet another Bull supercomputer will be part of PRACE's high performance computing infrastructure, contributing to an innovative experiment", declared Michel Guillemet, VP Innovative Products at Bull.