World’s most energy-efficient supercomputer recognized
Washington, November 21 (ANI): The high-performance computer QPACE (QCD Parallel Computing on the Cell) has been recognized as the most energy-efficient supercomputer in the world.
QPACE is at the head of the Green500 list, which provides a global ranking of energy-efficient supercomputers.
For a long time, in the world of supercomputers, performance was solely associated with speed.
This notion led to the development of computers that use enormous amounts of energy. Energy efficiency usually was ignored.
Not until the advent of increased discussions about the scarcity of natural resources and energy over the past years did this aspect gain in importance for the development of supercomputers.
Along with the Top500 list of the fastest computers, the Green500 list of supercomputers with the least energy use emerged as criterion for the rating and ranking of computer performance.
The QPACE supercomputer comes in at place 110 on the TOP500 list and has a computing power of 55 teraflop/s.
QPACE was developed by an academic consortium of universities and research centers as well as the German IBM research and development center in Boblingen within the framework of a state-sponsored research association.
It was deployed mid 2009 with four racks each at the Research Center Julich and at the University of Wuppertal.
It is being used for the simulation of fundamental forces in elementary particle physics, especially in the research area of quantum chromodynamics (QCD).
The heart of QPACE is the IBM PowerXCell 8i processor, an enhancement of the Cell/B.E. processor, which originally was developed by Sony, Toshiba and IBM for the Sony PlayStation 3.
With its nine processor cores, the chip can carry out a very large number of calculations simultaneously and at a high speed.
The novel concept of QPACE consists of connecting processors by a network of programmable units, called Field Programmable Gate Arrays (FPGA), to an efficient scalable computer.
Each of the QPACE installations in Julich and Wuppertal can reach a maximum performance of 100 TeraFlops (double precision).
That equates to 100 trillion (100,000,000,000,000) computing operations per second.
As a result of the scalability of the network, it is in principle possible to increase the performance to the PetaFlops scale (one quadrillion operations per second).