Brett Daniel Shehadey
Special Contributor for In Homeland Security
The latest Chinese supercomputer, Tianhe-2 (Milky War 2), tops the list of the world’s fastest supercomputers, replacing the US DOE “Titan” (the custom Cray XK7), clocking around 33 versus 17 petaflops. One petaflops is one quadrillion floating-point operations per second.
The petaflop threshold was achieved in 2008. It is a race of computation, largely between the US, Japan, China, and the EU. Incredibly fast calculations are not only important for a nation’s cryptography, encryptions systems, but also increasingly important for research, and Big Data in the informational world.
The Tianhe-2 was two years ahead of schedule; originally expected in 2015. Ironically, that was the date that the US wanted an exaflops computer—a machine one thousand times faster! Just last year the DARPA wanted to break the petascale altogether sometime by 2018, a three year difference from the previous target development goal. One exaflops equals one thousand petaflops (or one quintillion flops).
Supercomputer challengers have the power to affect almost everything in society as those abilities scale down in size. Take the Sandia Labs ASCI RED which came out in 1997 and reached 1.8 teraflops; almost ten years later, Sony, of Japan, came out with a PlayStation 3, in 2006, that ran at speeds of 1.8 teraflops. PlayStation 3 GPUs power supercomputers today (e.g. the USAF’s Condor Cluster in 2010).
If it is just a race of speed, then one might expect Japan, the EU and the US would be clear loser here with the Tianhe-2. But there is more to it than speed.
At least for the EU and the US, there are many more supercomputers currently in use. In 2010 the US was using 56% of the world’s supercomputers and china was only using 4.8% (Source: Top 500 Supercomputers). At present, according to the same source, the US share of world supercomputers is about 54% and China’s doubled to around 10 percent. Japan’s share of supercomputers has almost remained constant. China is also expected to rapidly increase their numbers in the long-term.
Performance capabilities of states have also seen a similar pattern—US and EU advantage declines overtime and China’s rises. Yet the architecture of supercomputing will also fundamentally change and this is even more important than just considering speed or number of computers in use.
American companies like IBM take the lead in the experimental cognitive computers—moving from sequential parallel-processing, clusters and silicon to—nanotechnology, virtual, holographic and DNA hard drives; “neuro-snyaptic cores,” and other potentially revolutionary leaps in brain mapping to computer projects. Hence those important fuzzy innovation variables that remain both elusive and difficult to gauge are also determinants of a clear victor.
Comments are closed.