INTERCONNECT AND THERMAL SYSTEM RELIABILITY FOR PROCESSOR UNITS
Authors: Dr. Paul P.E. Wang, Julia Purtell, and Michael Mil Company: Microsoft Corp Date Published: 1/17/2006
Pan Pacific Symposium
Abstract: Increasing requirements on computational speed and graphically virtual reality for game console are pushing the processor units toward the denser the routing layout and the higher the output wattage. The thermal interconnect of Power supply-Fan-Heat sink-Thermal Interface Material- Processor Unit is the key element in the effectiveness of the thermal reliability system. In this study the thermal performance of the cooling fan is assessed through the measurement of volumetric airflow rate and pressure drop and then used as targets for statistic process control and management. Thermal degradation of the Thermal Interface Material (TIM) due to pump out is studied by measuring the junction to heatsink thermal resistance while the interconnect system is under thermal aging. The end seal integrity of integrated Heatsink and heatpipe heatpipe (IHSHP) is verified by checking the Leadfree SnBi Solder bond leakage and loss of thermal conductivity due to factors such as thermal stress and fin surface contamination and corrosion. Accelerated life testing is conducted on all devices with multiple stress conditions in order to derive acceleration factor (AF) and life prediction to the user environment for each individual device. For system life prediction, the governing parameters of distribution model and life & stress relationship are derived in the process first. Then, the reliability integration of the whole thermal system will be conducted by using Reliability Block Diagram (RBD) to derive the life of the thermal system. Finally reliability allocation and device quality improvement are rationalized by considering the technical feasibility and cost structure in the near future. On the Ongoing Reliability Test (ORT) basis, the derived AF and one of the stress conditions are used in the ALT to ensure the life and fundamental mechanism of the device is maintained under the management of quality control chart and statistic run rule. A methodology is derived to correlate field failure data to the distribution of DVT to improve the accuracy of the life prediction model.