CHINA HAS THE FASTEST SUPERCOMPUTER TIANHE-2 IN THE WORLD FOR THIRD TIME
https://impnotice.blogspot.com/2014/06/china-has-fastest-supercomputer-tianhe.html
Zipping along at 33.86 petaflops (quadrillions of calculations per second), China's Tianhe-2 is the world's fastest computer for the third year in a row, according to Top500, a twice yearly compilation of the fastest computers on Earth.
The results were announced on Monday at the International Supercomputing Conference in Leipzig, Germany.
The Tianhe-2 was developed by China's National University of Defense Technology. Its predecessor, the Tianhe-1A, held the top spot on the supercomputer list in 2010. Top500 organizers note that the top 10 supercomputers are mostly unchanged from the last list, with the exception of a Cray XC30 computer, which is "installed at an undisclosed U.S. government site," according to a release on the Top500.org site.
Overall, the United States has by far the most supercomputers out of the top 500 systems, with China in a distant second; the UK, France and Germany lag far behind.
Supercomputers are not like the computers you use at home or in the office; they're not even much like the servers you’ll find in the office basement. They’re comprised of thousands of nodes and designed for crunching data and solving very difficult problems — in other words, research and development. That's why you'll find them hidden away at military facilities and laboratories.
Top500 notes that the rate of performance growth for the entire list is slowing down. Performance at the low end of the list typically grew at 90% year-over-year, but has now dropped to a 55% growth rate. Part of this deceleration is attributed to the dearth of new systems at the top. Despite that lack of change, these supercomputers will get more super. Tianhe-2 is projected to reach 100 petaflops by 2018.
Of the 500 systems on the list, 37 offer performance above the 1 petaflops range, nearly all of them use multicore processors, and 85% of them are running Intel processors. As for the systems themselves, 36% are built by HP and 33% by IBM. Cray builds 10%.
This is the 21st year that organizers in Germany and the United States have compiled the list.
The computer scientists measure system performance with the Linpack Benchmark, which is used in each system to solve "a dense set of equations." The benchmark is not a measure of the system's overall performance, but rather its performance on that task. Top500 organizers insist that no single benchmark can effectively measure the overall performance of a given system.
Have something to add to this story? Share it in the comments.
The results were announced on Monday at the International Supercomputing Conference in Leipzig, Germany.
Supercomputers are not like the computers you use at home or in the office; they're not even much like the servers you’ll find in the office basement. They’re comprised of thousands of nodes and designed for crunching data and solving very difficult problems — in other words, research and development. That's why you'll find them hidden away at military facilities and laboratories.
Top500 notes that the rate of performance growth for the entire list is slowing down. Performance at the low end of the list typically grew at 90% year-over-year, but has now dropped to a 55% growth rate. Part of this deceleration is attributed to the dearth of new systems at the top. Despite that lack of change, these supercomputers will get more super. Tianhe-2 is projected to reach 100 petaflops by 2018.
Of the 500 systems on the list, 37 offer performance above the 1 petaflops range, nearly all of them use multicore processors, and 85% of them are running Intel processors. As for the systems themselves, 36% are built by HP and 33% by IBM. Cray builds 10%.
This is the 21st year that organizers in Germany and the United States have compiled the list.
The computer scientists measure system performance with the Linpack Benchmark, which is used in each system to solve "a dense set of equations." The benchmark is not a measure of the system's overall performance, but rather its performance on that task. Top500 organizers insist that no single benchmark can effectively measure the overall performance of a given system.
Have something to add to this story? Share it in the comments.
