Abstract The rapid expansion of online platforms and web-based AI has driven a surge in computational power demand and data center growth, necessitating advanced cooling solutions to manage increasing power densities and energy consumption. This study evaluates the effectiveness of a two-phase direct-to-chip liquid cooling system in high-density rack environments. A key focus is the deployment of a 160 kW Refrigerant-to-Liquid (R2L) Coolant Distribution Unit (CDU), integrated with row and rack manifolds and server cooling loops. To emulate real-world thermal loads, Thermal Testing Vehicles (TTVs) equipped with 2.5 kW heaters covering 2500 mm2 are used to replicate server loads of up to 10 kW. The study details the system's design, commissioning, and extensive hydraulic and thermal testing, emphasizing pressure drops across supply manifolds, cooling loops, and return manifolds. Each Cooling Loop (CL) incorporates four Cold Plates (CPs) to optimize heat dissipation. Experimental data from TTV heating demonstrate the efficiency of two-phase cooling in transferring heat to chilled water using R134a refrigerant. At a volumetric flow rate of 0.48 LPM/kW, a Tcase as low as 56.4 °C was achieved, highlighting the system's capability to enhance thermal management for high-power-density chips. These findings validate two-phase cooling as a viable, energy-efficient solution for future data centers, addressing rising computational demands while promoting sustainable thermal regulation strategies.
Heydari et al. (Mon,) studied this question.