Dunking a computer into hot oil sounds a little reckless, but Intel has shown that servers can work more efficiently when they spend several months submerged this way.
The company recently finished a yearlong test of a liquid-cooling technique where servers slide into a custom server rack filled with mineral oil. The system, made by Austin, Texas-based Green Revolution Cooling, can cut power consumption from servers by 10 percent to 15 percent and from cooling equipment by about 90 percent, according to the company.
The server rack replacement can hold off-the-shelf servers and, because the oil doesn’t conduct electricity, components aren’t shorted out. The oil is circulated through the tank and cooled through a heat exchanger.
Intel is evaluating a number of radical new cooling techniques, including other liquid-cooling ideas, to better understand pathways to dramatically reduce energy consumption, says Mike Patterson, a senior power and thermal architect there. Data centers are big consumers of energy, which is significant to the overall cost of operation. In some cases, companies can’t even get enough power to run—and cool—the equipment properly.
The chip giant likes the immersive cooling idea so much it is looking at making servers designed especially for it. To work in liquid, servers need to have fans removed, hard drives encased, and a few other modifications. Patterson says that servers designed for immersive cooling would have a heat sink designed for liquid cooling (rather than air), no fans, and solid-state drives. In its tests, Intel found that the Green Revolution Cooling added only 2 to 3 percent of the server’s energy load for cooling, which is very low.
Green Revolution Cooling’s technology is particularly well suited for high-performance computing where servers are placed close together to coordinate on complex calculations. That server density means a high power load for the servers and cooling. Cooling can be half of the energy used in data centers, although the most efficient data centers spend about 10 percent of their total energy on cooling.
The three-year-old startup’s first customers tried immersive cooling not only to improve energy efficiency but also improve the reliability of components, says Andy Price, the company’s director of business development. Because some processors, including GPUs, were running so fast and hot, they suffered higher failure rates. Operating in mineral oil allows servers to operate at lower temperature, he says.
The return on investment for a high-performance computing lab is typically less than two years, Price says. The company is pursuing customers in cloud computing data centers as well.
Liquid cooling has been around for years. Cray supercomputers used liquid cooling and IBM has a system to pipe water directly next to server processors to wick away heat. (See “Hot Water Helps a Super-Efficient Supercomputer Keep its Cool.”) But liquid cooling is getting more attention as data center operators struggle to get sufficient power and reduce their energy load, particularly for high-performance supercomputers.
The main drawback to the immersive oil method is getting people used to it, says Patterson. Instead of just walking over to a server and popping in a new memory module, for example, data center professionals need to pull the server out of the tank, let the oil drain, and make the change cleanly.
“It’s truly a different operational mindset. You have to think differently,” he says.
Meta has built a massive new language AI—and it’s giving it away for free
Facebook’s parent company is inviting researchers to pore over and pick apart the flaws in its version of GPT-3
The gene-edited pig heart given to a dying patient was infected with a pig virus
The first transplant of a genetically-modified pig heart into a human may have ended prematurely because of a well-known—and avoidable—risk.
Saudi Arabia plans to spend $1 billion a year discovering treatments to slow aging
The oil kingdom fears that its population is aging at an accelerated rate and hopes to test drugs to reverse the problem. First up might be the diabetes drug metformin.
Yann LeCun has a bold new vision for the future of AI
One of the godfathers of deep learning pulls together old ideas to sketch out a fresh path for AI, but raises as many questions as he answers.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.