Jackson Braz Marcinichen

Advances in Electronics Cooling

This article highlights the advantages of on-chip microchannel cooling technology, based on first- and second-law analysis and experimental tests on two types of cooling cycles, the first driven by an oil-free liquid pump and the second by an oil-free vapor compressor. The analysis showed that the drivers of the fluid were the main culprits for major losses. It was further found that when energy recovery is of importance, making use of a vapor compression cycle increases the quality of the recovered energy, hence increasing its value. This was demonstrated by analyzing the synergy that can exist between the waste heat of a data center and heat reuse by a coal-fired power plant. It was found that power-plant efficiencies can be increased by up to 6.5% by making use of a vapor compression cycle, which results not only in significant monetary savings, but also in the reduced overall carbon footprints of both the data center and the power plant.

Green Cooling of High Performance Microprocessors: Parametric Study Between Flow Boiling and Water Cooling

Due to the increase in energy prices and spiralling consumption, there is a need to greatly reduce the cost of electricity within data centers, where it makes up 50% of the total cost of the IT infrastructure. A technological solution to this is using on-chip cooling with a single-phase or evaporating liquid to replace energy intensive air-cooling. The energy carried away by the liquid or vapour can also potentially be used in district heating, as an example. Thus, the important issue here is “what is the most energy efficient heat removal process?” As an answer, this paper presents a direct comparison of single-phase water, a 50% water ethylene glycol mixture and several two-phase refrigerants, including the new fourth generation refrigerants HFO1234yf and HFO1234ze. Two-phase cooling using HFC134a had an average junction temperature 9 to 15˚C lower than for single-phase cooling, while the required pumping power for the CPU cooling element for single-phase cooling was on the order of 20-130 times higher to achieve the same junction temperature uniformity. Hot-spot simulations also showed that two-phase refrigerant cooling was able to adjust to local hot-spots because of flow boilings dependency on the local heat flux, with junction temperatures being 20 to 30˚C lower when compared to water and the 50% water-ethylene glycol mixture, respectively. An exergy analysis was developed considering a cooling cycle composed by a pump, a condenser and a multi-microchannel cooler. The focus was to show the exergetic efficiency of each component and of the entire cycle when the subject energy recovery is considered. Water and HFC134a were the working fluids evaluated in such analysis. The overall exergetic efficiency was higher when using HFC134a (about 2%) and the exergy destroyed, i.e. irreversibilities, showed that the cooling cycle proposed still have a huge potential to increase the thermodynamic performance.

Towards development of a passive datacenter cooling technology: On-server thermosyphon cooling loop under dynamic workload

Despite many advances in electronics liquid cooling, air still remains one of the main means of cooling of high heat flux servers of datacenters. Regardless their long history of use, air-cooled blade servers continue to introduce strong nuisances that need to be considered during their implementation and operation, for instance: large energy consumption, difficult to manage the flow of air mastered by fans, and thus non-optimal spatial layout of components within a blade, high cost of air flow equipment, acoustical noise limitations, dust, etc. To overcome such problems, a new essential datacenter infrastructure is required.

Modeling of Two-Phase Evaporative Heat Transfer in Three-Dimensional Multicavity High Performance Microprocessor Chip Stacks

Three-dimensional (3D) stacking of integrated-circuit (IC) dies increases system density and package functionality by vertically integrating two or more dies with area-array through-silicon-vias (TSVs). This reduces the length of global interconnects and the signal delay time and allows improvements in energy efficiency. However, the accumulation of heat fluxes and thermal interface resistances is a major limitation of vertically integrated packages. Scalable cooling solutions, such as two-phase interlayer cooling, will be required to extend 3D stacks beyond the most modest numbers of dies. This paper introduces a realistic 3D chip stack along with a simulation method for the heat spreading and flow distribution among the channels of the evaporators. The model includes the significant sensitivity of each channels friction factor to vapor quality, and hence mass flow to heat flux, which characterizes parallel two-phase flows. Simulation cases explore various placements of hot spots within the stack and effects which are unique to two-phase interlayer cooling. The results show that the effect of hot spots on individual dies can be mitigated by strong interlayer heat conduction if the relative position of the hot spots is selected carefully to result in a heat load and flow which are well balanced laterally.

Two-phase mini-thermosyphon electronics cooling, Part 4: Application to 2U servers

This paper is the fourth part of the present study on two-phase mini-thermosyphon cooling. As mentioned in the first three parts, gravity-driven cooling systems using microchannel flow boiling can become a long-term scalable solution for cooling of datacenter servers. Indeed, the enhancement of thermal performance and the drastic reduction of power consumption together with the possibility of energy reuse and the inherent passive nature of the system offer a wide range of solutions to thermal designers. While Part 1 presented the first-of-a-kind low-height microchannel two-phase thermosyphon test results and Parts 2 and 3 showed the system scale steady and dynamic modeling and simulation results associated with this design using our inhouse simulator, Part 4 deals here with an end-user application, i.e. the cooling of a 2U server. The dynamic code of Part 3 is used to model the behavior of a mini-thermosyphon that would fit within the height of a 2U server (8.9cm high), while respecting the other geometric constraints (positions of the processors, distance of the processors to the back of the blade, etc.). Thus, the simulated system consists of two parallel multi-microchannel evaporator cold plates on the top of two chips of about 11cm2, a riser, a common water-cooled micro-condenser at the back of the blade, a liquid accumulator and a downcomer (including the piping branches to/from the two cold plates). First, an analysis of the steady-state operation highlights multiple solutions from which one is stable and one is unstable. Then, the influences of few parameters such as refrigerants, piping diameters, water coolant inlet temperature and flow rates, filling ratio and heat flux are evaluated. Simulations with unbalanced heat loads on the two chips being cooled in parallel then show the desirable flow distribution obtained in such gravity-driven systems. Finally, temporal heat load and water coolant flow rate disturbances are simulated and discussed. Noting all of these numerous influences on optimal mini-thermosyphon operation, the need for a accurate and detailed simulation code, benchmarked versus actual system tests, is seen to be imperative for attaining a good, reliable, robust design.

Two-phase mini-thermosyphon electronics cooling, Part 1: Experimental investigation

Efficient, small, state-of-the-art passive cooling two-phase systems, i.e. advanced micro-thermosyphon cooling systems, are viable solutions for high performance datacenter servers and power electronics cooling applications. The objective of this study is to push through the “two-phase threshold” that seems to be hindering the application of this cooling technology by offering here proven experimental results (Part 1), validated steady-state and transient simulation tools (Parts 2 and 3) and a server case study (Part 4). The experimental investigation in Part 1 presents the thermal-hydraulic performance of a mini-thermosyphon loop with a small riser height, Hriser = 15.0 cm. The thermosyphon loop has a multi-microchannel copper evaporator, mounted on top of a pseudo-chip CPU emulator (heat source). Experimental results for R134a, acquired under both pumped flow and passive thermosyphon driven flow (for direct comparison) for mass flow rates up to 10 kg/hr, uniform heat fluxes, q of up to 61.4 W/cm2 and refrigerant filling ratios up to 83% were obtained. An innovative thermal calibration method, developed as a non-intrusive mass flow measurement technique, has also been implemented to monitor the thermosyphons operation. Summarizing in brief, the two-phase thermosyphon loop with an integrated in-line liquid accumulator offered a very sustainable cooling performance for the microchannel/pseudo-CPU package, and is a first step forward in our effort towards the integration of such two-phase passive cooling devices for data center servers and other electronic devices at heat flux of up to 80 W/cm2 (or more).

Two-Phase Flow Control of Electronics Cooling With Pseudo-CPUs in Parallel Flow Circuits: Dynamic Modeling and Experimental Evaluation

On-chip two-phase cooling of parallel pseudo-CPUs integrated into a liquid pumped cooling cycle is modeled and experimentally verified versus a prototype test loop. The systems dynamic operation is studied since the heat dissipated by microprocessors is continuously changing during their operation and critical heat flux (CHF) conditions in the microevaporator must be avoided by flow control of the pump speed during heat load disturbances. The purpose here is to cool down multiple microprocessors in parallel and their auxiliary electronics (memories, dc/dc converters, etc.) to emulate datacenter servers with multiple CPUs. The dynamic simulation code was benchmarked using the test results obtained in an experimental facility consisting of a liquid pumped cooling cycle assembled in a test loop with two parallel microevaporators, which were evaluated under steady-state and transient conditions of balanced and unbalanced heat fluxes on the two pseudochips. The errors in the models predictions of mean chip temperature and mixed exit vapor quality at steady state remained within +/- 10%. Transient comparisons showed that the trends and the time constants were satisfactorily respected. A case study considering four microprocessors cooled in parallel flow was then simulated for different levels of heat flux in the microprocessors (40, 30, 20, and 10 W cm(-2)), which showed the robustness of the predictive-corrective solver used. For a desired mixed vapor exit quality of 30%, at an inlet pressure and subcooling of 1600 kPa and 3 K, the resulting distribution of mass flow rate in the microevaporators was, respectively, 2.6, 2.9, 4.2, and 6.4 kg h(-1) (mass fluxes of 47, 53, 76 and 116 kg m(-2) s(-1)) and yielded approximately uniform chip temperatures (maximum variation of 2.6, 2, 1.7, and 0.7 K). The vapor quality and maximum chip temperature remained below the critical limits during both transient and steady-state regimes.

Two-phase mini-thermosyphon electronics cooling, Part 3: Transient modeling and experimental validation

This paper is the third part of the present study on two-phase mini-thermosyphon cooling systems. As mentioned in the first two parts, gravity-driven cooling systems using microchannel flow boiling are a very promising long-term viable solution for electronics cooling and more specifically for datacenter servers. Indeed, the enhancement of thermal performance and the drastic reduction of power consumption together with the possibility of energy reuse and the inherent passive nature of the system offer a wide range of solutions to thermal designers. In order to design this new type of cooling system, a new novel simulation code specifically developed for this purpose is required. While Part 2 dealt with a steady-state simulation code, the present Part 3 considers the dynamic nature of the system. The dynamic simulator is a set of connected partial differential equations (temporal and spatial) solved for the four components of the thermosyphon, meaning for the micro-evaporator, riser, condenser and downcomer. Thermal inertia of the electronic package and role of a liquid accumulator are also accounted for. Predicted steady states obtained for 6 different heat fluxes (from 15.2 to 33.1 W/cm2) are compared to experimental results obtained with the test loop presented in Part 1 in terms of chip temperature and system pressure. Mean errors of 2.9 and 3.1% are respectively found and good performances of the heat transfer prediction methods used in the simulator are emphasized. Additionally, the dynamic response to a heat load disturbance is compared with experimental results in terms of chip temperature. Two variations with different time constants are both observed experimentally and predicted numerically. Finally, the predicted mass flow rate variations are discussed.

Two-phase mini-thermosyphon electronics cooling, Part 2: Model and steady-state validations

In the present study, a simulation code specifically developed to evaluate the thermal-hydraulic performance of thermosyphon cooling loops is validated through the experimental results obtained in the Part 1. It considers levels of heat load conventionally observed in real servers of datacenters, which means idle, normal and maximum clock speed of actual microprocessors. The thermosyphon is a very compact unit with a height of 15 cm and capable of safely operating up to a heat flux of 80 W cm-2. The loop basically is comprised of a riser, a downcomer, a micro-evaporator and a counter flow tube-in-tube condenser. The latter is cooled by cold water whose mass flow rate can be controlled through an external pump (speed control), so that parameters such as saturation temperature and/or condenser outlet subcooling can be adjusted for a pre-defined set point, and thus increasing the range of operability of the cooling loop. Other parameters were also explored experimentally, cooling looping overall performance, chip (junction) temperature, whilst the critical heat flux was estimated from a leading CHF method. Finally, the study showed that the passive two-phase closed loop thermosyphon cooling system is a safe and energetically viable technology solution for the next generation of datacenters.

Passive Thermosyphon Cooling System for High Heat Flux Servers

The main aim of the current paper is to demonstrate the capability of a two-phase closed thermosyphon loop system to cool down a contemporary datacenter rack, passively cooling the entire rack including its numerous servers. The effects on the performance of the entire cooling loop with respect to the server orientation, micro-evaporator design, riser and downcomer diameters, working fluid, and approach temperature difference at the condenser have been modeled and simulated. The influence of the thermosyphon height (here from 5 to 20 cm with a horizontally or vertically oriented server) on the driving force that guarantees the system operation whilst simultaneously fulfilling the critical heat flux (CHF) criterion also has been examined. In summary, the thermosyphon height was found to be the most significant design parameter. For the conditions simulated, in terms of CHF, the 10 cm-high thermosyphon was the most advantageous system design with a minimum safety factor of 1.6 relative to the imposed heat flux of 80 W cm−2. Additionally, a case study including an overhead water-cooled heat exchanger to extract heat from the thermosyphon loop has been developed and then the entire rack cooling system evaluated in terms of cost savings, payback period, and net benefit per year. This approximate study provides a general understanding of how the datacenter cooling infrastructure directly impacts the operating budget as well as influencing the thermal/hydraulic operation, performance, and reliability of the datacenter. Finally, the study shows that the passive two-phase closed loop thermosyphon cooling system is a potentially economically sound technology to cool high heat flux servers of datacenters.Copyright