An official website of the United States government
The increasing prevalence of high-performance computing data centers necessitates the adoption of cutting-edge cooling technologies to ensure the safe and reliable operation of their powerful microprocessors. Two-phase cooling schemes are well-suited for high heat flux scenarios because of their high heat transfer coefficients and their ability to enhance chip temperature uniformity. In this study, we perform experimental characterization and deep learning driven optimization of a commercial two-phase cold plate. The initial working design of the cold plate comprises a fin height of 3mm, fin thickness of 0.1 mm, and a channel width of 0.1 mm.A dielectric coolant, Novec /HFE 7000, was impinged into microchannel fins through impinging jets. A copper block simulated an electronic chip with a surface area of 1˝ × 1˝. The experiment was conducted with three different coolant inlet temperatures of 25◦ C, 36◦ C, and 48◦ C with varying heat flux levels ranging from 7.5 to 73.5 W cm2. The effects of coolant inlet temperatures and flow rate on the thermo-hydraulic performance of the cold plate were explored. In two-phase flow, increasing coolant inlet temperature results in more nucleation sites and improved thermal performance consequently. Thermal resistance drops with flow rate in single-phase flow while it is not affected by flow rate in nucleate boiling region. An improvement in the design of the cold plate was carried out, with the goal of increasing the number of bubble sites and flow velocity at the root fins, by cutting the original fins and creating channels perpendicular to the original channels. Three design parameters, fin height, width of machined channels, and height of short fins preserved through machined channels, were defined. It was observed that widening the machined channels and cutting fins to some point can improve the thermal performance of the cold plate. However, removing fins excessively adversely affects the thermal performance of the cold plate because of loss of heat transfer surface area. Moreover, preserving the short fins through the machined channels decreases thermal resistance as they increase heat transfer surface area and nucleation sites. Furthermore, a deep learning-based compact model is demonstrated for the two-phase cold plate design in the specific range of geometry and flow conditions. The developed compact model is utilized to drive the single and multi-objective optimization to arrive at global optimal results.
more »
« less
This content will become publicly available on February 1, 2026
Thermal challenges in heterogeneous packaging: Experimental and machine learning approaches to liquid cooling
In recent years, electronic packaging has evolved significantly to meet demands for higher performance, lower costs, and smaller designs. This shift has led to heterogeneous packaging, which integrates chips of varying stack heights and results in non-uniform heat flux and temperature distributions. These conditions pose substantial thermal management challenges, as they can create large temperature gradients, which increase thermal stress and potentially compromise chip reliability. This study explores single-phase liquid cooling for multi-chip modules (MCMs) through a comprehensive experimental and machine learning approach. It investigates the impact of chip spacing, height, fluid flow rate, fluid inlet location, and heat flux uniformity on chip temperature and the thermohydraulic performance of a commercial cold plate. Results show that increasing coolant flow from 1 LPM to 2 LPM decreased thermal resistance by 26 %, with heat losses remaining below 5 %. The left inlet configuration improved temperature uniformity compared to the right, though both yielded comparable thermal performance. Adjusting heater spacing impacted temperature distribution based on inlet position, and lowering one heater by 1 mm raised its temperatures by 15 ◦C due to increased thermal resistance from thermal interface material. A transient test demonstrated the cold plate’s quick response to power surges, in which there is only a 1 ◦C spike above steady state. Complementing these findings, an Artificial Neural Network (ANN) model was developed with optimized architecture specifically for the unique challenges of this study. The ANN model was rigorously validated using an independent dataset, achieving highly accurate temperature predictions (R2 = 0.99) within 2.5 % of experimental
more »
« less
- Award ID(s):
- 2209776
- PAR ID:
- 10611011
- Publisher / Repository:
- Elsevier - Applied Thermal Engineering
- Date Published:
- Journal Name:
- Applied Thermal Engineering
- Volume:
- 260
- Issue:
- C
- ISSN:
- 1359-4311
- Page Range / eLocation ID:
- 125081
- Subject(s) / Keyword(s):
- Multi-chip module Liquid cooling Machine learning Neural network Heat sink Data centers Heat transfer Thermal management
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The variety of new electronic packaging technologies has grown significantly over the last 20 years as a result of market demands for higher device performance at lower costs and in less space. Those demands have pushed for heterogeneous packaging, where computer chips with different stack heights are closely packed, creating nonuniform heat flux and temperature and additional challenges for thermal management. Without implementing an appropriate thermal management strategy for heterogeneous packages, large temperature gradients can be observed within the package, which would increase the thermal stresses on the chip and raise reliability issues. To mimic this real-life scenario of such packaging, an experimental setup was designed and built. The design of the new experimental setup consists of four identical 1.2 cm × 1.2 cm ceramic heaters, each of which is connected to a separate power supply and can reach a heat flux of 140 W/cm2. Accordingly, this mock package is capable of delivering different power levels to mimic different multicore microprocessor conditions. To give the heater the ability to move precisely in the x-, y-, and z-directions, each heater is mounted to an XYZ linear stage. Deionized water (DI) was used as the working fluid, and a pin-fin heat sink was used to run the initial steady-state tests on the experimental rig. The tests showed how different flow rates at a constant fluid temperature and input power affect the temperatures of the heaters and the thermohydraulic performance of the heat sink. In addition, a three-dimensional numerical model has been developed and validated with experimental data in terms of heat sink pressure drop and the temperatures of the heaters.more » « less
-
Miniaturization and high heat flux of power electronic devices have posed a colossal challenge for adequate thermal management. Conventional air-cooling solutions are inadequate for high-performance electronics. Liquid cooling is an alternative solution thanks to the higher specific heat and latent heat associated with the coolants. Liquid-cooled cold plates are typically manufactured by different approaches such as: skived, forged, extrusion, electrical discharge machining. When researchers are facing challenges at creating complex geometries in small spaces, 3D-printing can be a solution. In this paper, a 3D-printed cold plate was designed and characterized with water coolant. The printed metal fin structures were strong enough to undergo pressure from the fluid flow even at high flow rates and small fin structures. A copper block with top surface area of 1 inch by 1 inch was used to mimic a computer chip. Experimental data has good match with a simulation model which was built using commercial software 6SigmaET. Effects of geometry parameters and operating parameters were investigated. Fin diameter was varied from 0.3 mm to 0.5 mm and fin height was maintained at 2 mm. A special manifold was designed to maximize the surface contact area between coolant and metal surface and therefore minimize thermal resistance. The flow rate was varied from 0.75 L/min to 2 L/min and coolant inlet temperature was varied from 25 to 48 oC. It was observed that for the coolant inlet temperature 25 oC and aluminum cold plate, the junction temperature was kept below 63.2 oC at input power 350 W and pressure drop did not exceed 23 Kpa. Effects of metal materials used in 3D-printing on the thermal performance of the cold plate were also studied in detail.more » « less
-
More than ever before, data centers must deploy robust thermal solutions to adequately host the high-density and high-performance computing that is in high demand. The newer generation of central processing units (CPUs) and graphics processing units (GPUs) has substantially higher thermal power densities than previous generations. In recent years, more data centers rely on liquid cooling for the high-heat processors inside the servers and air cooling for the remaining low-heat information technology equipment. This hybrid cooling approach creates a smaller and more efficient data center. The deployment of direct-to-chip cold plate liquid cooling is one of the mainstream approaches to providing concentrated cooling to targeted processors. In this study, a processor-level experimental setup was developed to evaluate the cooling performance of a novel computer numerical control (CNC) machined nickel-plated impinging cold plate on a 1 in. 1 in. mock heater that represents a functional processing unit. The pressure drop and thermal resistance performance curves of the electroless nickel-plated cold plate are compared to those of a pure copper cold plate. A temperature uniformity analysis is done using compuational fluid dynamics and compared to the actual test data. Finally, the CNC machined pure copper one is compared to other reported cold plates to demonstrate its superiority of the design with respect to the cooling performance.more » « less
-
This paper focuses on two-phase flow boiling of dielectric coolant HFE 7000 inside a copper multi-microchannel heat sink for high heat flux chip applications. The heat sink is composed of parallel microchannels, 200 μm wide, 2500 μm high, and 20 mm long, with 200-μm-thick fins separating the channels. The copper heat sink consists of almost 100 channels connected by a longitude groove with a nearly trapezoidal cross section. Coolant impinges down to the base at the groove and then goes along the microchannels. A copper block heater arrangement was used to mimic a computer chip with a footprint of 1”x1” (6.45 cm2). The base heat flux was varied from 7.75 W/cm2 to 96.1 W/cm2 and the mass flux from 547.6 to 958.4 kg/m2s, at a nominal saturation temperature of 54 °C. Heat transfer coefficients as high as 57.5 kW/m2K were reached, keeping the base temperature under 66 °C with a maximum of 21.9 kPa of pressure drop, for inlet subcooling of 5 degree and a coolant flow rate of 958.4 kg/m2. Effects of inner diameter of tubing on thermal performance and pressure drop are also discussed. It was observed that an increase of tubing inner diameter by 60 % can result in increase of heat transfer coefficient by 47.8 % and reduction in pressure drop by 63 %.more » « less
