Skip to Content

Hot Water Helps a Super-Efficient Supercomputer Keep Its Cool

IBM’s water-cooling system lowers Leibniz SuperMUC supercomputer energy use by 40 percent.

When it comes to supercomputers, it’s not all about raw horsepower. Keeping a lid on energy consumption has become one of the biggest technical challenges for making beefier and faster supercomputers.

A computer rendering of the new supercomputer. Credit: IBM.

The Leibniz Supercomputing Centre today announced that its “SuperMUC” supercomputer (MUC is the airport code for the nearby Munich airport), built in conjunction with IBM, consumes 40 percent less energy than a comparable machine. The trick to dramatically lowering energy consumption is water cooling, an approach IBM continues to pursue to pack more computing power into smaller packages.

Computer water cooling systems circulate water over active components, such as processors and memory, to wick away heat. Often, these cooling systems have to lower the temperature of incoming water before it’s circulated, according to the Leibniz Supercomputing Centre. But the SuperMUC can operate with hot water up to 40°C (over 100°F). That means the outgoing hot water can be cooled by ambient air or through a heat exchange, rather than compressor cooling equipment.

In the Leibniz installation, water heated by the supercomputer will be circulated to a heat exchanger and then used to keep the building warm during the winter time. The Centre estimates it will save about one million Euros each year this way. For all its energy efficiency though, SuperMAC is no slouch in the computing department: equipped with more than 150,000 processor cores in 9,400 compute nodes and more than 300 terabytes of RAM, it’s the fastest supercomputer in Europe with a peak performance of three petaflops, or 10,000,000,000,000,000 Floating Point Operations per second.

The basic computer cooling technology, called Aquasar, stems from IBM Research’s Zurich lab. Small pipes, called microchannels, carry water directly to the surface of server processors. Once water is cooled off, it can be recirculated to remove heat again.

IBM’s x iDataPlex dx360 M4 server cooling system. Credit: IBM.

Work on water cooling has been going on for years but in a typical data center or supercomputer, the temperatures of servers and other equipment are controlled by passing cool air over racks. Cooling represents roughly half of the energy consumption from these large computing centers.

An energy-efficient computing architecture is important for environmentally-conscious Germany, says Herbert Huber from the Leibniz Supercomputing Centre. “Germany is very greenish so energy and power awareness is a crucial aspect,” he said.

In general, techniques for managing energy consumption and peak power are becoming more important as designers seek to pack more powerful computers into smaller spaces. Because of its energy-efficient design, SuperMAC is ten times more compact and has better peak performance than a comparable air-cooled machine, according to the Leibniz Supercomputing Centre.

IBM expects hot water cooling to be central to future supercomputers, including three-dimensional chip architectures where multiple processors are stacked on top of each other. Researchers are working on tiny tubes only 50 microns across embedded between the processor stacks, according to IBM Research. With water cooling, IBM predicts that by 2025 a supercomputer could fit into the casing of a 2012 desktop computer.

The water cooling system sits on top of SuperMUC. Credit: IBM

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

The problem with plug-in hybrids? Their drivers.

Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

How scientists traced a mysterious covid case back to six toilets

When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.