Invited Talk: Pathfinding to Maximization of AI/HPC/GPC System Performance

This paper will deliver the pathfinding to maximize AI/HPC/GPC system performance from a thermo-mechanical prospective. In this article, the challenges that emerge to develop continuous innovations for manufacturable products will be explained. These recommendations utilize the opensource ecosystem standard since sticking to a standard that does not evolve rapidly can impact the design freedom The package architecture, intended application, system level, and manufacturing solution impact on the performance and system IT load balancing are also taken into account. To address these challenges and to push the boundary, a strategy/roadmap and a pathfinding will be presented. Furthermore, this study will be disclosing key factors which need to be considered in developing the present and next generation of thermo-mechanical solutions. These factors include enabling and working on/around the manufacturing infrastructure, accommodating future heterogeneous integration, utilizing current/future infrastructure of board level manufactured assembly, addressing the solution from the system level, extending air-cooling, enabling liquid-cooling, and considering immersion liquid-cooling. Based on these factors, a pathfinding strategy will be built on four elements/pillars which are hosting facilities, driving ecosystems, defining cost based on operating condition and reliability, and synergistic management of workload in both IT equipment and cooling systems.