China's Tianhe-2 Supercomputer Still has the Most Flops (Japan's K-Computer's in 4th place)
Word's Fastest Computer
Reports from the SC13 Supercomputer Conference in Dever, CO, U.S.A. have once again confirmed that China's Tianhe-2 supercomputer is leading the world's petaflop count - it pulls off 33.86 of them. Tianhe-2 has held this title since June of this year. That's cool, though aside from making the top of SC13's list, nobody really knows what the computer's actual purpose and utility will be.
What's a petaflop, one might wonder? Well, basically it's the measure we now must use to quantify a supercomputer's benchmarked capability, i.e., Tianhe-2 throws down 33.86 of them per second, and that's just under twice as many as #2. We didn't always use petaflops though, we actually only passed the teraflop barrier in 2008 with IBM's Roadrunner.
For all practical everyday people and their purposes, there's no need to trot out the exponents and hypersyllabic numeric values - all you really need to know, and feel free to share this with friends and family, is that "China has the world's fastest supercomputer because it can do a kabillion jillion calculations per second."
At 10.51 petaflops, Fujitsu's K-Computer was the world's fastest for a period in 2011. The same machine, managed by the RIKEN Advanced Institute for Computational Science (AICS), now sits at #4.
Who's Really Winning at Supercomputing?
America. By a quite a bit, actually. Of the world's top 10, the United States has #2, #3, #5, #7, and #9. And it's not just speed. While the Cray XK7 Titan only turns 17.59 petaflops, it does so with approximately 1/5 of the processing cores and less than half the energy required by Tianhe-2. The U.S. also uses supercomputers to do actual stuff like climate modeling, finding cats on the internets, and other massively complicated huge-data tasks - and that counts for a lot.
What's Clear is This:
Computer's are getting a helluva lot faster at a helluva lot faster rate. Considering that along with this: as reported earlier this month, some of Google's data center systems are exhibiting what is arguably - though speculatively - emergent, semi-biological patterns that engineers actually can't completely explain. Might be a good idea to start thinking about how to be nice to your machines.
If we do not, if we're not careful, someday - sooner than most would guess - supercomputers might decide to power humans down.
• • •