Aquarius HPC Cooling System Yields Impressive Energy Efficiency Results After NREL/Sandia Test
Technical Report Reveals How the Cold-Plate Liquid-Cooling Design Performed
May 21, 2019
After a performance evaluation, the National Renewable Energy Laboratory (NREL)—in partnership with Sandia National Laboratories—determined the Aquarius fixed-cold-plate liquid-cooling design for server racks provides a very high percentage of heat capture direct to water—up to 98% when evaluating compute nodes only.
"We found that a very high percentage of heat was captured direct to water," the report reads. "Based on observations, it seems possible to increase the supply IT water temperature to the compute nodes (tests were conducted using 85F supply water). This would allow even warmer facility supply as well as push the return temperatures up, which is ideal for waste heat reuse and/or use of dry coolers for on-site water savings."
As announced in 2018, the labs were testing the Aquarius cooling system, created by Aquila, at NREL's High-Performance Computing (HPC) Data Center in the Energy Systems Integration Facility. Aquila and Sandia chose NREL's HPC Data Center for the initial installation and evaluation because the data center is configured for warm water liquid cooling, which is an energy-efficient approach because water is able to absorb and remove much more heat energy than air per unit of volume. This allows us to provide adequate cooling while using less energy than fans would require.
The Aquarius system is designed to provide energy-efficient touch cooling to off-the-shelf dual-processor server systems. The fixed cold plates are arranged in such a way that water flows between them with a pattern that ensures a consistent temperature across their entirety.
NREL ran the tests in its award-winning, ultra-energy-efficient data center to evaluate the cooling system. This set of components, named Yacumama, held 36 compute nodes. The Aquarius system requires no fans, so all the power pull from the compute nodes is used for actual computing, not to spin fans. Through testing and after almost 10 months of operation, the cluster has required no maintenance.
The primary objective of the evaluation was to determine the quantity of energy being dissipated to water to determine how close the fixed cold-plate approach is to a room-neutral solution. The team tested in two ways—one test to learn the range of reasonable changes one can make to the system, and a long-duration test.
The team completed its analysis of the Aquarius cooling system with suggestions for further work, which mainly included proposed procedural practices and a couple minor hardware modifications.
Learn more about the Aquarius cold-plate cooling system in the technical report.