Optimisation of Liquid Heat Exchangers for CPU Thermal Management

Abstract

The research presented in this thesis addresses critical gaps in the understanding related to electronic cooling with conjugate heat transfer effects, and its integration into practical Central Processing Unit (CPU) architectures, culminating in the development of an automated optimisation framework for liquid jet array impingement heat exchanger design. The work investigates the intricate relationship between heat transfer performance, hydraulic efficiency, and practical constraints, offering novel insights into the design of advanced, energy-efficient cooling solutions tailored to the demands of modern and future electronics. The first stage of analysis focused on the efficiency of heat spreading in Integrated Heat Spreaders (IHS). While heat spreading efficiency has parallels with fin efficiency in traditional systems, its specific implications for concentrated heat sources with convectively cooled IHSs had not been thoroughly explored. A Genetic Algorithm-based framework was employed to optimise the distribution of convective heat transfer coefficients across the IHS surface. The results demonstrated that non-uniform heat transfer distributions significantly enhanced thermal performance, reducing peak temperatures and minimising gradients across the concentrated heat source. This approach highlighted the potential of using targeted liquid cooling strategies to optimise the conjugate heat transfer, providing a new perspective on managing concentrated heat sources. Next, the practical application of these theoretical insights to more real-world liquid cooling systems was investigated. An automated multi-objective optimisation workflow was developed to generate high-performance impinging jet array designs optimised for thermal and hydraulic objective functions. Considering CPU-type heat sources with IHSs, the research confirmed that distributed liquid cooling strategies, tailored to specific operational conditions, achieved superior performance, balancing cooling intensity with energy efficiency. This work underscored the practical feasibility of implementing distributed cooling strategies in commercial applications and further validated the effectiveness of the proposed methodology in achieving optimised cooling solutions. Finally, the influence of real-life CPU architecture on thermal management strategies was explored. The optimisation framework was adapted to a commercial CPU package, the Intel i5-12600K, to investigate the unique thermal challenges posed by modern high-powered processors. By incorporating discrete and high heat flux cores on the underside of the semiconductor die into the simulation-based optimisation workflow, the research demonstrated that approximately 30% of the total source-to-sink thermal budget could be attributed to the heat transfer in the silicon die. This underscored the importance of considering design architectural influences when designing liquid cooling systems. Experimental verification under realistic operating conditions showed strong agreement between simulation and measured performance, reinforcing the viability and practicality of the developed simulation design approach. Overall, this research provides a comprehensive framework for optimising liquid jet array impingement heat exchangers, bridging theoretical advancements with practical applications in electronics cooling. By addressing the complexities of heat spreading efficiency, distributed convective cooling, and CPU architectural influences, this research culminates in a robust and experimentally verified simulation design tool for tackling the evolving thermal challenges associated with high-powered CPUs. By validating key hypotheses and providing practical solutions, the findings contribute to the broader field of electronics cooling, offering pathways to more efficient, reliable, and sustainable cooling solutions.