Graduation Semester and Year
Fall 2024
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Mechanical Engineering
Department
Mechanical and Aerospace Engineering
First Advisor
Dereje Agonager
Second Advisor
Abdolhossein Haji-Sheikh
Third Advisor
Miguel A. Amaya
Fourth Advisor
Amir Ameri
Fifth Advisor
Satyam Saini
Abstract
This dissertation investigates advanced thermal management strategies for high-performance computing systems to address the escalating thermal challenges posed by increasing power densities and computational demands. Chapter 1 evaluates the efficacy of forced convection Single-Phase Immersion Cooling (SPIC) with a dielectric fluid (EC-110) compared to traditional air cooling for a 776 W server. Computational Fluid Dynamics (CFD) simulations demonstrate significant reductions in chip junction temperatures, maximum Dual In-line Memory Module (DIMM) temperatures, and server pressure drop with immersion cooling, highlighting its superior heat dissipation capabilities.
Chapter 2 delves into the comparative analysis of forced and natural convection Single-Phase Immersion Cooling techniques. The study examines the impact of varying inlet temperatures and server configurations on component temperatures, pressure drop, and pumping power. Results indicate that forced convection systems generally outperform natural convection systems in maintaining lower component temperatures, even with enhanced cooling characteristics in the latter.
Chapter 3 focuses on optimizing heatsink designs for immersion-cooled servers to minimize case temperatures while adhering to pumping power constraints. A multi-objective multi-variable optimization approach, utilizing CFD simulations and a Design of Experiments framework, explores the influence of fin count, thickness, and height on heatsink performance. The study identifies optimal design parameters for both CPU and GPU heatsinks, leading to improved thermal performance and reduced pressure drop.
Finally, Chapter 4 explores the potential of Electrochemical Additive Manufacturing (ECAM) in fabricating high-performance cold plates for cooling high-powered chips with non-uniform power distributions. CFD simulations and experimental validation demonstrate that ECAM-enabled cold plates with intricate internal flow structures exhibit superior thermal performance compared to conventional designs, offering significant advancements in thermal management for next-generation electronics.
This dissertation provides valuable insights into the design, optimization, and evaluation of advanced cooling technologies, including immersion cooling, optimized heatsinks, and novel manufacturing ECAM technology, contributing to the development of more efficient and sustainable high-performance computing systems.
Keywords
Data Center Cooling; Thermal Management, Electrochemical Additive Manufacturing, Immersion Cooling
Disciplines
Engineering | Heat Transfer, Combustion
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Gupta, Gautam, "Advanced Thermal Management Solutions for High-Powered Chips: A Comprehensive Study on Immersion Cooling and Electrochemical Additive Manufacturing-Based Cold Plates" (2024). Mechanical and Aerospace Engineering Dissertations. 424.
https://mavmatrix.uta.edu/mechaerospace_dissertations/424
