News, Science & Technology

IBM’s Sequoia Now World’s Fastest Supercomputer

Note: This article is hosted here for archival purposes only. It does not necessarily represent the values of the Iron Warrior or Waterloo Engineering Society in the present day.

On June 14th, 2012, IBM announced that it had developed the world’s fastest supercomputer claiming the title from the Fujitsu K Computer in Japan.

The term supercomputer refers to any computer at the bleeding edge of current processing power, which is usually measured in calculation speed. The earliest supercomputers used only a few cores, but in 2012 consist of massive setups of thousands of processors.

The Sequoia uses BlueGene/Q servers and achieved 16.32 petaflops (or 16.32 quadrillion floating point operations per second) at the Lawrence Livermore National Laboratory using over 1.5 million processor cores and 1.6 petabytes of memory. The BBC likens its calculation power in one second to “what otherwise would take 6.7 billion people using hand calculators 320 years.” In comparison, the Fujitsu K Computer achieved 10.51 petaflops with over 705000 processor cores.

Although the Sequoia draws about 7.9MW of power during operation, it is also one of the most energy-efficient supercomputers in existence – a third more efficient than the K Computer, in fact, which consumes 12.6MW during operation. Only 50% of the average data centre’s energy consumption and carbon footprint stems from powering the necessary cooling systems. The Sequoia uses a hot-water cooling system. Coolant temperatures near active components such as the processors and memory modules in the Sequoia reach up to 45 degrees C. Water has a higher heat capacity than air, and thus can absorb more heat from the system. Although more effective and efficient than air-cooled systems, water-cooled systems have not gained widespread use in personal computers because of leakage risks.

Like most American supercomputers, the IBM Sequoia is classified as “dual-use” by the United States, meaning that it has both civilian and military applications. The primary purpose of the Sequoia will be in nuclear weapons modelling, specifically in hydrodynamics and properties of materials at the extreme pressures and temperatures that would ensue after a nuclear attack. Furthermore, the Sequoia will peform statistical anaylsis to support the life extensions of existing, aging weapons systems, as the US has not created any new nuclear weapons since the end of the Cold War, in 1991. On the other hand, the US has not conducted any underground nuclear testing since 1992 either, after adequate computing power rendered the need for real-life testings almost completely obsolete.

Possible civilian applications of the Sequoia include modelling climate change, the human genome, or the flow of blood through an artificial heart valve.

However impressive IBM’s accomplishment in the Sequoia may be, however, modern-day tech historians would do well to remember Moore’s Law. Moore’s law states that the number of transistors that can be placed inexpensively on an integrated circuit doubles every two years. Intel executive David House predicted that the period for doubling chip performance would be 18 months. What this means is that you shouldn’t expect the Sequoia to sit at the top of the supercomputer heap for too long before it too is surpassed by a new system. Mark my words, we’ll hit the 1-exaflop barrier before the decade is through.

Leave a Reply