Parametric Analysis of Thermo-Mechanical Behavior for HBM in System-In-Package Configurations

The main purpose of this study is to comprehensively analyze thermo-mechanical behavior according to various conditions and variables when HBM is mounted on SIP. Nowadays, HBM (High Bandwidth Memory) device is considered to be the most important part for implementing AI (Artificial Intelligence) calculation for ChatGPT or LLM (Large Language Model) users. SIP (System in Package) refers to the system assembly where a set of HBM devices are equipped with GPU(Graphic Processor Unit) by connecting signals via TSV (Through Silicon Via) and Silicon IP (Interposer), which differs depending on the customer's requirements. Due to the inherent thin thickness of HBM chips and thermal compression process at stacking step, it highly requires to quantify thermomechanical behavior that causes stress concentration. It was verified that the stress of local regions in HBM devices could be increased as thermal gradient in SIP becomes large. In addition, a parametric study on the influence of distance between GPU and HBMs on the stress concentration were also carried out. This study presents the usefulness of FE simulation for predicting the structure robustness in advance for System in Package of HBM devices prior to the customer's reliability test with high fidelity.