ENERGY | WIRELESS | NANOTECH | MEMS | OPTICS | QUANTUM | 3D | CHIPS | ALGORITHMS

Tuesday, April 09, 2013

#CHIPS: "Budget Supercomputing Made Easy"

Today more and more servers are turning into budget supercomputers, often called high-performance computers (HPCs), a teraFLOP at a time by merely plugging in an Intel Xeon Phi coprocessor. The task then becomes parallelizing their algorithms, for which many tutorials are available since the Xeon Phi is just another multi-core x86-based processor. Now in addition a new "best practices" treatment just for the Xeon Phi has just been made available free by the Partnership for Advanced Computing in Europe (PRACE): R. Colin Johnson @NextGenLog


PRACE member, the Poznan Supercomputing and Networking Center (Poland) currently has 2048 Intel Xeon cores in its SGI Altix UV 100 shared memory cluster called “Chimera” and can upgrade to more cores by adding Xeon Phi coprocessors at any time.
Further Reading