Hybrid Two-Phase Cooling Technology For Next-Generation Servers: Thermal Performance Analysis

This paper advances the work presented at ITHERM 2020 where a novel passive two-phase technology was introduced for more efficient cooling of datacenters compared to air-based cooling solutions (e.g. In-Room, In-Row, etc.). The proposed hybrid two-phase cooling technology uses passive low-height thermosyphons to dissipate the heat generated by high power components inside the servers. The heat is then transferred to multiple rack-level thermosyphons, equipped with one or more overhead compact condensers, which in turn dissipate the total heat from the rack into a room-level water-based or air-based cooling loop. Air-cooling is still used for the low heat-generating components (e.g. memory units, secondary chips, etc.) to make the cooling technology cost effective. This study is focused on a thermal performance analysis of the hybrid two-phase cooling technology applied to a single server (2U HP ProLiant DL180 G6), where low-height thermosyphons are installed to cool the two CPUs, and the remaining server components are cooled via air in forced convection using server-level fans. The thermal performance of the low-height thermosyphons is extracted from previous experimental measurements carried out at Nokia Bell Labs and presented at ITHERM 2020. Additional measurements considering the effect of the refrigerant charge on the thermal resistance and associated chip case temperature are conducted. The air-cooling performance and flow distribution are investigated via CFD simulations performed in COMSOL that characterize the conjugate heat transfer between server components and cooling air. The experimental and numerical results presented here show that, at the maximum server IT load of 587 W, the hybrid two-phase technology ensures sufficient cooling for all the hardware components while significantly reducing the energy consumption of the fans. Specifically, the fan contribution is reduced from 24% (traditional air-cooling) to about 2% of the server effective IT load (hybrid two-phase cooling), demonstrating the potential of the proposed technology to scale existing hardware and build next-generation datacenters, while keeping low costs and being environmentally-friendly.

Fundamental study on microlayer dynamics in nucleate boiling

Lubomír Bures

Dynamics of nucleate boiling are strongly affected by the formation and behaviour of the microlayer, a layer of liquid underneath growing bubbles. As a result of its minute thickness, very high heat fluxes occur within the microlayer and its evaporation contributes significantly to the overall heat transfer. Microlayer formation is, however, not guaranteed and the transition from the contact-line to the microlayer regime of nucleate boiling is not fully understood. The difficulty of experimental investigation of the microlayer and the uncertainties surrounding its formation and subsequent evolution motivate the use of Direct Numerical Simulation (DNS) to model its behaviour. In this work, a computational strategy for utilising DNS to model nucleate boiling by resolving explicitly the microlayer is developed. The numerical method is based on the resolution of continuum conservation equations for incompressible two-phase flows in Cartesian and axisymmetric cylindrical coordinates. The phasic interface is tracked by means of the geometric Volume-of-Fluid (VOF) method and the algorithm is applicable both to adiabatic and volatile flows. Online, implicit coupling of the fluid and solid domains for the solution of the conjugate heat-transfer problem is included and closure models for the treatment of the interfacial heat-transfer resistance and the dynamic contact angle are introduced. A rigorous verification and validation exercise is performed to evaluate the efficacy of the numerical algorithm. Subsequently, a theoretical criterion for modelling the transition between contact-line and microlayer regimes is derived, and tested. Very good agreement is found with the reference experimental and simulation data and the predictive power of the criterion is demonstrated with the aid of DNS results. The computational procedure is then validated for simulations of nucleate boiling with resolved microlayer using relevant experimental data recently measured at the Massachusetts Institute of Technology; it is shown that the main observed growth features and surface heat-transfer characteristics are well-reproduced using the overall method. A sensitivity study of the dependence of the initial microlayer thickness on the growth conditions is performed and a universal equation describing the thickness distribution is proposed with liquid properties and bubble expansion rate being the governing parameters. Finally, the computational method is extended to coarse-mesh problems by introducing several reduced-order models and the full bubble-growth cycle from nucleation to detachment is simulated. Good agreement with reference measurements is again achieved and the experimental findings regarding the force balance during nucleate boiling are confirmed.

EPFL2021

Dynamic Modeling and Experimental Evaluation of a Controlled Two-Phase On-Chip Cooling System Designed for High Efficiency Datacenters

Nicolas Lamaison

An important challenge for the next generation of high performance datacenters is to guarantee sufficient, reliable, green and cheap cooling of the server blades. Since this seems unrealistic with air cooling technologies due to the expected ever higher chip heat fluxes (in excess of 100 W/cm2), a novel more effective cooling system is required. Switching to two-phase on-chip cooling systems would decrease the energy consumption, enhance the thermal performance and allow the reuse of the extracted heat. However, the design of such systems seems nontrivial because of the lack of simulation tools able to reproduce their behavior. In this context, the present thesis describes the one-dimensional dynamic modeling of an on-chip two-phase liquid pump cooling system for multiple micro-processors. The simulation code includes state-of-the art two-phase flow methods, two-dimensional conduction in the heat sink, heat flux dependent mass flow rate distribution between parallel micro-evaporators, and micro-condenser and in-line liquid accumulator models. Three different scales of simulations, namely the heat sink, multiple parallel heat sinks and entire system scales, are presented. An experimental test bench with a cooling capacity of about 300W is used to validate the code. Mass flow rate distribution, energy oriented performance maps, chip temperature profile, heat flux disturbances under controlled conditions and charge influence are studied experimentally. The experimental data are satisfactorily reproduced by the simulations with, for example, 266 steady-state mean chip temperatures, obtained for balanced and unbalanced heat fluxes conditions, predicted within ± 2K. The most significant discrepancy between simulations and experiments is found to be the pressure transient prediction. The validated code is then used to simulate a compact two-phase on-chip cooling system designed for a real datacenter blade server. At the heat sink level, start-up and hot-spots are studied using 2D temperature maps and time-strip analysis. Furthermore, the flow distribution between the parallel micro-evaporators of the studied layout is shown to be unique and thermal-hydraulic maps are generated. The system start-up and operation under random and periodic heat load disturbances with different control strategies are then analyzed. Finally, the thermo-economic integration of a datacenter with a district heating network is investigated.

EPFL2014

State of the Art of High Heat Flux Cooling Technologies

Bruno Agostini, Bruno Michel, Jung Eung Park, John Richard Thome, Leszek Wojtan

The purpose of this literature review is to compare different cooling technologies currently in development in research laboratories that are competing to solve the challenge of cooling the next generation of high heat flux computer chips. Today, most development efforts are focused on three technologies: liquid cooling in copper or silicon micro-geometry heat dissipation elements, impingement of liquid jets directly on the silicon surface of the chip, and two-phase flow boiling in copper heat dissipation elements or plates with numerous microchannels. The principal challenge is to dissipate the high heat fluxes (current objective is 300 W/cm2) while maintaining the chip temperature below the targeted temperature of 85°C, while of second importance is how to predict the heat transfer coefficients and pressure drops of the cooling process. In this study, the state of the art of these three technologies from recent experimental articles (since 2003) is analyzed and a comparison of the respective merits and drawbacks of each technology is presented. The conclusion is that two-phase flow boiling in microchannels is the most promising approach; impingement cooling also has good prospects but single-phase liquid cooling is probably only a short-term solution. As an example of the state of the first technology, the Heat and Mass Transfer Laboratory at Ecole Polytechnique Fédérale de Lausanne has already achieved 200 W/cm2 of cooling in a first prototype, with a low pumping power, good temperature uniformity, and at the required maximal operating temperature.