Exascale Systems Boggle Minds and Budgets

Can the Department of Energy create a supercomputer so massive that even it cannot fathom it?

The human mind boggles (well, this human’s mind boggles, anyway) at how high numbers can go. When trying to wrap my head around how “super” existing supercomputers are and projected supercomputers can be, I wanted to get an idea of the difference between what we have now and what we could potentially have by 2018; not even the infallible Internet wanted to give me a definite answer.

Exascale Systems Boggle Minds and Budgets
Photo by Argonne National Laboratory

Wikipedia, the unbending voice of accuracy (har, har) mocked my efforts by telling me that “quadrillion” isn’t even one particular number! Instead, I was presented with this jibber jabber:

Quadrillion may mean either of the two numbers (see long and short scales for more detail):

  • 1,000,000,000,000,000 (one thousand million million; 1015; SI prefix peta-) – for all short scale countries – increasingly common meaning in English language usage
  • 1,000,000,000,000,000,000,000,000 (one million million million million; 1024; SI prefix: yotta-) – for all long scale countries – increasingly rare meaning in English language usage

Oh. Thanks for clearing that up. You think Mr. T would accept that answer? He pities the fool. I, on the other hand, was driven to choose history as a major. We all deal with these things as we see fit.

Recently, the Obama administration set aside a 2012 budget proposal of $126 million (that’s an unfathomably big number, too!) toward the U.S. Department of Energy and its development of an exascale system — the next generation to follow our current petaflop (a thousand trillion floating point operations per second) supercomputers. This is in contrast to the comparatively paltry $24 million allotted to the DOE in 2011.

The Tianhe-1A, a Chinese supercomputer, is currently the fastest, and exascale systems are projected to be a thousand times more powerful — with one million trillion calculations per second — by 2018. If successful, the edge the US would have over the rest of the world in scientific research (including the study of climate change, energy production, nuclear science, and biology) and national security could be pretty staggering.

Exascale, shmexascale. Petaflop, shmetaflop. In the world of science, as in just about anything else, you get what you pay for — and all too often, bigger is better.