China has moved ahead of the US with the world’s fastest supercomputer – half the size, lower power and 50 percent faster than America’s best effort.
The Chinese National University of Defence Technology has designed a system that has reached record speeds of 2,507 petaflops, or two quadrillion calcs per second, in the National Supercomputer Centre in Tianjin.
The completion of the Tianhe-1A project marks a significant step in the technological rivalry between the two superpowers. Although the official list of top 500 supercomputers will not be released until next week, it is almost inconceivable that China’s Tianhe-1A will not be perched at the top.
As mentioned by the NY Times this is not the first time America has been knocked from the top spot – Japan managed to lay claim to the fastest machine from 2002 until 2004 – however there are significant ramifications for the rivalry.
The supercomputers are used to enhance simulation and research in areas such as climate modelling, defence, and genomics, and such projects are crucially able to attract top researchers from across the world.
Ironically it is US-headquartered firm Nvidia which is powering the supercomputer with its Tesla M2050 GPUs.
The record breaking system utilises 7,168 Tesla GPUs, alongside 14,336 CPUs. Tesla GPUs are the fundamental reason why the system is able to use so little space and energy while generating so much power. To put it into perspective it would take 50,000 CPUs to produce the processing speed of Tianhe-1A, taking up twice the amount of floor space.
Guangming Liu, chief of National Supercomputer Center in Tianjin noted how instrumental GPUs are to the supercomputer: “The performance and efficiency of Tianhe-1A was simply not possible without GPUs. The scientific research that is now possible with a system of this scale is almost without limits; we could not be more pleased with the results,” said Liu.