Datacenter Liquid Cooling Lab Evaluation

This proposed project aims to address the increasing demand for data centers driven by the rise of Artificial Intelligence (“AI”), machine learning, and cloud computing. In California, Pacific Gas and Electric Company (“PG&E”) estimates an additional three and a half (3.5) gigawatts (“GW”) of datacenter capacity within the next five years. Currently, traditional datacenters rely heavily on air cooling, primarily using computer room air conditioning (“CRAC”) units. Approximately thirty-three to forty percent (33-40%) of the electrical energy consumed by these datacenters is dedicated to cooling, presenting a significant opportunity to reduce overall energy consumption by twenty (20%) through targeted liquid cooling strategies for computer chips. Various liquid cooling technologies, such as single and two-phase immersion cooling, single and two-phase direct liquid cooling, and hybrid strategies that include liquid and air cooling, are currently available on the market.To determine the most promising cooling technologies for energy efficiency programs and to understand the barriers to their adoption, this project will involve experimental benchtop testing of three targeted cooling solutions for one-unit (“1U”) servers with identical computing specifications. Chip performance and operating temperatures will be monitored under varying workloads, with ITUE (defined below) and cooling power consumption recorded and analyzed for comparison. Testing will take place over a period of three (3) months to assess the long-term reliability of each cooling technology. These findings will serve as experimental validation for performance metrics used in the CalNEXT sister project (ET25SWE0046), where they will be utilized to evaluate the energy efficiency gains from liquid cooling compared to air cooling, along with the challenges associated with retrofitting and technology adaptation. UC Davis Western Cooling Efficiency Center (“WCEC”) (the “Project Team”) will also conduct a scalability analysis, considering key metrics such as Power Usage Effectiveness (“PUE”), Total Energy Usage (“TUE”), IT Energy Usage (“ITUE”), and other relevant factors (https://knowledge.sdialliance.org/data-centermetrics#903dbdc8aae84211aa42aa33c625ab7e) to measure the potential for widespread adoption of these cooling technologies in California.