The most recent Top500 list of fastest supercomputers around the world shows little change among the 10 leading machines. China still holds the top spot in what amounts to a petty competition among governments for bragging rights (a high percentage of the leading machines are run either by government agencies or universities—that is, de facto government agencies). David Schneider reports at IEEE Spectrum, “Indeed, there is little change in the top 10 positions, with only two new names appearing in that elite group.... As you’d expect with such little turn over at the top of the list, the overall rate of growth in performance of the world’s top supercomputers has been slowing in recent years.”
Part of the slowdown in progress is no doubt a result of a declining Moore’s Law: the economics of chip making may now be such that packing more transistors on a chip is no longer cost effective. But other factors are likely playing a role, including the challenge of moving data quickly among all the processors in the machine. “But the summed performance of the world’s 500 fastest supercomputers is still up by 55 petaflops/s over the TOP500’s June 2015 ranking,” Schneider added.
A far more interesting version of the Top500 list would be those machines whose funding comes purely from private sources—that is, machines whose value is recognized and backed voluntarily rather than through the coercive power of taxation. Such machines would represent not only processing capability, but a real demand for that capability—as well as funding sources that build the economy rather than tearing it down. In the meantime, however, the pervasive notion that governments should have a major role in funding science will continue, and one of the results will be numerous impressive machines of dubious value to the society that is forced to fund them.
Read more about supercomputers