Will We See ExaScale Computing this Decade?

Stagnating speed increases among the world’s most powerful supercomputers leads some to believe that Exascale computing might miss its 2020 mark.

supercomputer, tianhe, exaflop, petaflop, simulation, scienceLast week Top 500 released its list of the world’s fastest, most-powerful computers. For the third consecutive time China’s Tianhe-2 has topped the list. Capable of performing 33.86 quadrillion computations per second (22.86 petaflops) the Tianhe-2 is nearly twice as fast as its closest rival. While those numbers are certainly impressive they hide an ominous trend – for the second consecutive time the trend in computation power has grown at a slower rate than expected.

For the past five years computer engineers, institutional research directors and scientists have been designing, fundraising for and awaiting a quantum leap in computational ability – the exascale supercomputer.

One thousand times faster than a petaflop, an exaflop-class supercomputer would signal a new era of computational ability. While initial hopes were that an exascale supercomputer may be developed sometime between 2018 and 2020, the TOP 500’s observed trend is disheartening news.

Today a number of teams worldwide are racing to build the first exaflop machine. Some are using enormous clusters of smaller, smartphone like chips to churn out calculations. Others are looking for better materials to continue the progress of their brute force schemes. Unfortunately, for seemingly every supercomputing method the same obstacle stands in their way: energy costs.

Supercomputing currently costs around $1M per petaflop of calculations per year. With systems continually scaling up, the price of high-performance computing is becoming a burden on computer development.

While the world of supercomputing may seem obscure, the overall performance of the world’s top 500 supercomputers does have real world repercussions. Super computers help solve problems relevant to engineering, material science, energy, Earth science and basic research. What’s more, researchers are hoping that with the help of supercomputers we’ll be able to create highly detailed maps of our cosmos as well as peer deep into the human brain with accurate simulations of its inner workings.

For those projects alone, research into supercomputer hardware and programming materials should be richly funded. Especially seeing as quantum computers are still far from ready for primetime.

Infographic of Top 500 June 2014 Results

Source: TOP 500