Nuclear fusion research receives a boostby Utilities ME Staff on Mar 31, 2014
The Helios supercomputer, which supports research work aimed at controlling nuclear fusion, has been boosted with the addition of more processing power.
Designed and operated by Bull, the Helios supercomputer installed in Rokkasho, Japan, is boosted by an additional 380 Intel Xeon Phi coprocessors to increase its power to almost 2 Petaflops.
The system provides modeling and simulation capacity which is open to all European and Japanese researchers under the ‘Broader Approach’, a research programme that complements the international cooperative ITER programme.
The new Intel Xeon PHI coprocessors that will be incorporated into Helios will enable researchers to take advantage of increased computing performance. Their massively parallel architecture delivering leading performance per watt, foreshadows technologies that will ultimately lead to Exa-scale computing.
“We are delighted to be helping the CEA and the community working on Nuclear Fusion to develop the knowhow and computing resources that will allow them to significantly expand the potential for research associated with the ITER programme,” said Pascal Barbolosi, Vice-President, Extreme Computing at Bull.
The architecture of the Helios supercomputer initially featured 4,410 bullx B510 compute nodes, with 8,820 Intel Xeon E5 processors producing a power of 1.5 Petaflops. The 180 new bullx B515 compute nodes will each include two Intel Xeon Phi coprocessors delivering an additional 400 Teraflops, taking the total power of the Helios system to almost 2 Petaflops.
“Intel is convinced that Exascale computing will represent a major technological advance for the scientific community working on Nuclear Fusion projects. The Intel Xeon Phi range of coprocessors has been designed with this in mind, and we are very pleased that the Helios supercomputer will be benefiting from this,” explained Stéphane Negre, CEO of Intel France and Regional Manager of Intel Western Europe.
The Center of Expertise for Parallel Programming – created by Bull in close cooperation with Intel and based at Bull’s offices in Grenoble, France – is closely involved in this project, providing training, porting and optimisation of computing codes dedicated to this vital area of energy research.