Researchers from the Western Cooling Efficiency Center have published a new study on a liquid-cooled modular data center design for high-density computing and artificial intelligence applications.
The study, published in the Journal of Electronic Packaging, presents a chiller-less cooling system that uses a single liquid loop to remove heat from server components and reject it to outdoor air. The design uses direct-to-chip liquid cooling, a MUSE cold plate, and a microchannel polymer heat exchanger dry cooler to support compact, high-compute-density data centers.
The modeling results show that, at an outdoor temperature of 40°C, optimized liquid and fan flow rates can reduce cooling power to 2.1% of compute power for a GPU case temperature of 75°C. The study also found that improvements to the thermal interface material and dry cooler had the largest effect on system performance, with the best-case cooling power falling below 1% of compute power.
The findings point to a potential path for reducing cooling energy in modular edge data centers, which are expected to play a growing role in AI, advanced sensing, medical imaging, autonomous systems, and other data-intensive applications.