As well as comprising many processors, supercomputers have been able to take advantage of other developments in computer technology. These include multiple cores on a single computer, and the powerful processing capabilities of graphics processing units (GPUs), which have far surpassed the speed of central processing units at the kinds of calculations important for scientific computation: namely, performing complex mathematical calculations in parallel. These incredible speeds have been in part driven by the video games industry and the demand for hyper -realistic, 3D graphics. While the video games consoles and computers in the average person’s home are leaps ahead of those around 10 or 15 years ago, supercomputers can make use of the equivalent of thousands of these to act together as one.
The K Computer in Japan – until June 2012 the world’s fastest supercomputer – has 88,128 processors, each with 8 cores, making 705,024 cores in total. The website for the K computer contains some useful comparisons:
“If all seven billion people in the world were gathered together and given a calculator to perform one calculation per second continuously for 24 hours per day without rest, it would take approximately 17 days to complete the number of calculations that the K computer is able to perform in just one second. If we think of the calculation speed of a regular computer as the speed of a snail, then the K computer has the speed of a jet airplane.”
How K was overtaken by Sequoia: the Linpack Benchmark
Supercomputers are measured for their ability to sustain high-speed calculations, namely a series of linear equations. In June 2012 Sequoia, an IBM Bluegene computer, reached 16.32 Petaflops (million Megaflops), a 55% increase on Japan’s K computer, making it the fastest in the world. It is also more energy efficient than the K computer – using 7.9 Megawatts to K’s 12.6 Megawatts