Abstract
This work examines the thermal solution needs of data center semiconductor equipment by analyzing cooling requirements in addition to the limitations of legacy approaches such as air-cooling. This research explores the feasibility of two-phase immersion cooling and its trajectory in future applications such as artificial intelligence (AI). The thermal performance investigation of the two-phase immersion cooling thermal management system in this study is tailored for high-performance Solid State Drives (SSDs) as they evolve to meet the growing demands of AI and edge computing. A numerical analysis was performed to investigate modeling a nano-porous coating as a mode of heat transfer enhancement for SSDs using computational fluid dynamics (CFD) software. The goal of this study is to develop an approach to simulating an approximation of a Boiling Enhancement Coating (BEC); hence, a User-Defined Function (UDF) was used as a method for modeling BECs. The results of the simulation analysis without UDF integration show that the boiling behavior of dielectric two-phase fluids, such as Fluorinert FC-72, differs considerably compared to distilled water as a heat transfer fluid. Moreover, the results show approximately a 3-4°C difference between surface and inner SSD temperature. With the integration of a UDF, the results show a 6% difference in total heat transfer after several seconds of boiling time vs the default built-in boiling correlations. Simulations using UDFs similarly show a 0.18˚C increase in average surface temperature over the same length of simulated boiling time when compared to the default. Experimental validation was conducted, and results showed good agreement between time-to-boil with an observed difference of 5.6%. Validation also showed a 2% difference in the average surface temperature of the SSD cover and the T3 thermocouple location during boiling. In addition, the delta in temperature rise between thermocouple locations T1 and T2 showed agreement in experimental testing and those observed in simulations, and were observed to be approximately 0.24˚C and 0.17˚C between simulation and experimental testing, respectively.