PARAM supercomputer among world’s most power efficient
02 Jan 2014
The Centre for Development of Advanced Computing (C-DAC) has achieved a significant milestone with its PARAM Yuva II supercomputer being ranked first in India, 9th in the Asia Pacific region and 44th in the world among the most power efficient computer systems as per the Green500 list announced at the Supercomputing Conference (SC'2013) in Denver, Colorado, USA.
Supercomputers, in general, consume a lot of electrical power and produce much heat that necessitates elaborate cooling facilities to ensure proper operation. This adds to increase in the total cost of ownership (TCO) of a supercomputer. To draw focus towards development of energy efficient supercomputers, Green500 ranks computer systems in the world, according to compute performance per Watt, thus providing a world ranking based on energy efficiency.
Energy consumed by supercomputers is measured at various Levels – L1, L2, L3 - for purpose of reporting. As the level increases, accuracy and rigor of measurement exercise also increases. ''It is also a measure of our capability and noteworthy that C-DAC is the second organisation worldwide to have carried out the Level 3 measurement of power versus performance for the Green500 List.''
Congratulating C-DAC for this significant achievement, minister for communications and information technology Kapil Sibal said supercomputing is very important for the all-round advancements in the country, and that the government was planning a big impetus for capacity building and advanced R&D in this area. He wanted C-DC to make many more contributions as a key player in the computing world.
PARAM Yuva-II uses hybrid technology – processor, co-processor and hardware accelerators - to provide the peak compute power of 520.4 Teraflops using 210 kilo Watt power. The interconnect network comprises of homegrown PARAMNet-III and Infiniband FDR system area network. This system is designed to solve large and complex computational problems. The system has 200 terabytes of high performance storage, and requisite system software and utilities for parallel applications development.
