IBM Xeon-Based Supercomputer To Hit Three Petaflops

SuperMUC, to be built for German researchers, will use hot-water cooling to cut power use by 40 percent

IBM is to build an Intel Xeon-based supercomputer for the Leibniz Supercomputing Centre (LRZ) in Garching, Germany, that will deliver up to three petaflops of performance and will rely on a hot water cooling system – allowing it to consume 40 percent less power than a comparable air-cooled machine, according to Intel.

The system, called SuperMUC, is set to go online in 2012 and will form part of the Partnership for Advanced Computing in Europe (PRACE) high-performance computing (HPC) infrastructure for European research and industrial institutions, Intel said.

Hot water makes for efficiency

LRZ's existing HLRB II supercomputer, SGI Altix 4700

IBM delivered its first hot water-cooled supercomputer, called Aquasar, to the Swiss Federal Institute of Technology Zurich (ETH Zurich) in July. That system uses waste heat from the supercomputer to provide heat to university buildings, reducing the computer’s carbon footprint by 85 percent.

The company’s hot-water cooling approach cools processors and other components with warm water of up to 60 degrees Celsius, using micro-channel liquid coolers attached directly to the processors.

Water typically removes heat 4,000 times more efficiently than air, Intel said.

SuperMUC will use energy-efficient processors and dynamic systems management to further reduce energy consumption.

“SuperMUC will provide previously unattainable energy efficiency along with peak performance by exploiting the massive parallelism of Intel’s multicore processors and leveraging the innovative hot water cooling technology pioneered by IBM,” said Dr. Arndt Bode, Chairman of the Board of Directors of LRZ, in a statement. “This approach will allow the industry to develop ever more powerful supercomputers while keeping energy use in check.”

The system will be based on the IBM System x iDataPlex and will use more than 14,000 Xeon processors. Intel compared its performance to that of more than 110,000 PCs.

Collaboration

The design work will be carried out by IBM’s development team in Boeblingen, which also helped build Aquasar. Intel is collaborating on the energy-efficient design, while LRZ will contribute its experience in operating high-end supercomputing systems – it already houses the HLRB-II, a giant Linux cluster from SGI.

The system will be used to test theories, design experiments and predict outcomes in areas such as cosmology and seismology. It is being funded jointly by the German federal government and the state of Bavaria.

In November Intel highlighted how HPC is entering the mainstream and is now used in everything from measuring blood flow in the human body and mapping the evolution of the ocean in light of climate change, to carrying out data analysis in the financial sector.

In August a team led by renowned astrophysicist Professor Stephen Hawking at the UK Computational Cosmology Consortium (COSMOS) in Cambridge, announced its choice of SGI’s Altix UV 1000 – described as the world’s fastest supercomputer – to support its research into the origins of space.

“Recent progress towards a complete understanding of the universe has been impressive, but many puzzles remain,” said Hawking at the time. “Cosmology is now a precise science, and we need supercomputers to calculate what our theories of the early universe predict and test them against observations of the present universe.”