Waterloo team cools chip hotspots with single coolant drops

Researchers at the University of Waterloo have developed a hybrid cooling system that uses tiny drops of specialized coolant to reduce the temperature of simulated computer chip hotspots by up to 17 degrees Celsius in just 15 seconds. This novel approach combines general immersion cooling with inexpensive deionized water and targeted spot cooling of especially hot areas with drops of costly coolant. The system is designed to address the significant heat generation in artificial intelligence data centers, where small areas on chips can become much hotter than others in a matter of seconds, potentially causing components to slow down, age faster, and undermine overall reliability.

The innovation stems from the work of a research team led by Dr. Sushanta Mitra, a professor of mechanical and mechatronics engineering at Waterloo and executive director of the Waterloo Institute for Nanotechnology. The team, which included postdoctoral researcher Dr. Tian-Yu Zhang and PhD student Arnav Banerjee, built upon previous work to encapsulate one liquid in a thin skin of another. By selecting an electronics coolant with a much lower boiling point than water, they created a solution that is also heavier than water, preventing it from mixing and allowing it to stay put when placed on a hotspot.

In laboratory testing, a single drop of this coolant boiled and reduced the temperature of a simulated computer chip hotspot submerged in a tank of water by up to 17 degrees Celsius in 15 seconds. This reaction occurred several times faster than conventional immersion cooling methods. Dr. Mitra explained that boiling soaks up a significant amount of heat, and the rising bubbles stir the surrounding water, carrying heat away more efficiently. This mechanism provides the thermal impact of boiling only where and when it is specifically needed, offering a precise response to transient thermal spikes.

To further refine the technology, the researchers used their encapsulation technique to wrap the coolant drops in a thin skin of silicone oil. They added tiny, heat-conducting ceramic particles to this coating to help control how quickly the drops boil. The division of labor in the system allows water to handle steady, everyday heat, while the specialized drops take care of sudden spikes, creating a more efficient and economical cooling strategy.

The research team envisions the system ultimately working in one of two ways. In the first scenario, coolant drops would circulate through the water and activate only when they hit hotspots. In the second scenario, drops would be delivered directly to hotspots detected by sensors.

The hybrid cooling approach has potential applications beyond data centers, extending to technologies including electric vehicles, batteries, high-powered lasers, and LED lighting. The study, titled “A hybrid immersion cooling assisted by liquid-liquid encapsulation: A game changer for thermal management of transient hotspots,” was recently published online in the Journal of Colloid and Interface Science.

By addressing the specific issue of transient hotspots, the researchers have provided a solution that could make cooling AI data centers cheaper and more efficient. The use of inexpensive deionized water for general cooling, combined with targeted application of costly coolant for hotspots, offers a balanced approach to managing heat.

Subject of Research: Hybrid immersion cooling system for AI data centers using encapsulated coolant drops

Article Title: Waterloo innovation could make cooling AI data centers cheaper and more efficient

Article References: Waterloo innovation could make cooling AI data centers cheaper and more efficient. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: AI data centers, thermal management, immersion cooling, University of Waterloo, chip hotspots, coolant encapsulation, electronic cooling, Journal of Colloid and Interface Science, Waterloo, innovation, make, cooling