Two-Phase Mini-Thermosyphon for Cooling of Datacenters: Experiments, Modeling and Simulations

Abstract

Nowadays, datacenters heat density dissipation follows an exponential increasing trend that is reaching the heat removal limits of the traditional air-cooling technology. Two-phase cooling implemented within a gravity-driven system represents a scalable and viable long-term solution for datacenter cooling in order to increase the heat density dissipation with larger energy efficiency and lower acoustic noise. The present article builds upon the 4-part set of papers presented at ITHERM 2016 for a 15-cm height thermosyphon to cool a contemporary datacenter cabinet, providing new test data over a wider range of heat fluxes and new validations of the thermal-hydrodynamics of our thermosyphon simulation code. The thermosyphon consists of a microchannel evaporator connected via a riser and a downcomer to a liquid-cooled condenser for the cooling of a pseudo-chip to emulate an actual server. Test results demonstrated good thermal performance coupled with uniform flow distribution for the new larger range of operating test conditions. At the maximum imposed heat load of 158 W (corresponding to a heat flux of 70 W cm−2) with a water inlet coolant at 20 °C, water mass flow rate of 12 kg h−1 and thermosyphon filling ratio of 78%, the pseudo mean chip temperature was found to be 58 °C and is well below the normal thermal limits in datacenter cooling. Finally, the in-house LTCM’s thermosyphon simulation code was validated against an expanded experimental database of about 262 data points, demonstrating very good agreement; in fact, the pseudo mean chip temperature was predicted with an error band of about 1 K.

Enhancing Two-Phase Cooling Efficiency through Thermal-Aware Workload Mapping for Power-Hungry Servers

“…The use of twophase cooling, rather than single-phase liquid-based cooling systems, is strongly motivated due to their reduced mass flowrates, lower pumping power, and smaller facility size [14], while providing higher heat transfer coefficients and more uniform temperature profiles [12]. Previous thermosyphon prototypes such as [15], however, had a very large footprint area (1m ⇥ 1m) making them impractical in commercial servers. Nonetheless, recent work by [16] and [8] led to the design of micro-scale thermosyphons which can be placed directly on top of a CPU.…”

Design of a Two-Phase Gravity-Driven Micro-Scale Thermosyphon Cooling System for High-Performance Computing Data Centers

“…The most recent thermosyphon built, which demonstrates comparable performance results to a pumped cooling loop system, has 15 cm height but a large footprint area of 1 m × 1 m [7], [8]. This mini-thermosyphon validates the two flow regimes: i) gravitational dominant regime, and ii) frictional dominant regime.…”

Modeling of a Two-Phase Thermosyphon Loop With Low Environmental Impact Refrigerant Applied to Electronic Cooling

“…This change of slope is characterized by the authors as a change in the flow pattern, possibly a transition from slug to annular flow. The model also had a change of slope, even though it appears that the predictability was not as accurate as for lower heat loads, perhaps pointing out that the model could be improved for different flow patterns.While contrasting the mass flow rate as a function of heat load, obtained by the model ofOng et al (2017) with their experimental data, it is evident that the model showed both gravitational and frictional dominant regimes across their whole heat load input range, as did the experimental campaign. However, the model was shown to overpredict the mass flow rate especially for lower heat fluxes (15 to 25 W/cm 2 ) which characterize the gravitational dominant regime.…”

“…The study performed by Ong et al (2017) centered on adding data for a wider range of imposed heat load for a thermosyphon loop of 15 cm of height, designed to cool a pseudo-chip which emulated a server. The authors performed more experimental tests and provided new validations for the LTCM code on working fluid's mass flow rate, mean chip temperature, and system pressure.…”

“…In addition to the lower energy consumption of the closed thermosyphon loop system, the exemption of mechanical components means a more reliable system. Without moving parts, it is possible to reduce costs, vibrations that could upset the electronic apparatuses and, consequently, reduce noise levels (Ong et al , 2017). The possible drawback of the thermosyphon loop system is regarding instabilities and flow reversal, so the system must be carefully designed for its whole range of operation, so as to avoid undesirable operational problems such as evaporator dry out, difficult start-up or condenser flooding (Agostini and Yesin, 2008).…”