Srinidhi’s supercomp third fastest
WASHINGTON: A supercomputer rigged up from off-the-shelf components by an Indian researcher at VirginiaTech has been ranked third in the TOP500 list of the world’s fastest machines released this weekend.
Srinidhi Varadarajan ‘s’ 1100-Apple Mac contraption cranked out more than 10 Teraflops (or trillion calculations per second) to place third behind the top-ranked Japanese Earth Simulator (35.86 TFlops) and the second-placed U.S. Department of Energy’s Los Alamos National Laboratory ASCI Q (13.88 TFlops), the gurus in the business announced at the annual supercomputing mela in Phoenix, Arizona.
The VirginiaTech achievement, reported in these columns last month, has been attracting wide attention at the jamboree. What was remarkable, the conference noted, was that Srinidhi and his team was able to build the system in less than four months for only $5.2 million – less than one tenth the average cost of comparable systems. Such a breakthrough could change the economics of supercomputing.
VirginiaTech’s supercomputer is also, by far, the fastest machine ever owned by an educational institution. Most supercomputers in the Top25 are owned by government labs specializing in high-end research.
The latest Top500 list also showed that the most powerful supercomputer in India is not at C-Dac, Pune, as it is commonly believed, but with Intel India in Bangalore, whose X-series cluster IBM Xeon’s ran at 1.1 teraflops to place at 105th. C-Dac’s Param Padma came in at 258.
Outside the western world and Japan, which placed at the top, the most powerful machine was at the Chinese Academy of Science, whose DeepComp 6800 placed 14th.
There were few other changes in the top 10 list, but the supercomputing wizards noted that the list of cluster systems in the TOP10 has grown impressively to seven systems.
Both the VirginiaTech machine and Param Padma are cluster systems i.e., they are built with workstations or PCs as building blocks and often connected by special high-speed internal networks. The number of clusters in the full TOP500 grew also again strongly, now totaling 208 systems up from 149 six months ago
Impressive!
