Analytical, Numerical and Experimental Study of Phase Change Material in TIM2 Application for High-Power Server CPUs

The thermal design power (TDP) of server CPU keeps increasing 10-15% generation to generation (G2G) and it pushes the TDP of server CPU to increase to 300+W. In overall thermal stack-up of the server in data center, thermal interface material (TIM) between the CPU and cooling solution is normally called TIM2 to distinguish itself from TIM1 between the die and CPU integrated heat spreader (IHS). There are a few types of TIM2 for this type of application for CPU, including phase change material, grease, thermal pad, and thermal film. Grease is one typical material used as TIM2 due to its relatively lower cost, but risk of pump out in usage is a concern. There have been progresses in PCM, including new phase change material w/ thermal conductivity from 2 to 7 W/k/m. Phase change material (PCM) based TIM2 material is usually characterized by a TIM2 tester. Test results are reported in terms of thermal impedance vs. bond line thickness (BLT) and/or averaged contact pressure. However, thermal performance of phase change material (PCM) collected from TIM2 tester often cannot represent its behavior in actual usage. The primary reason is because the two contacting surfaces typically have warpages that deviate significantly from the relatively flat surfaces used in TIM2 tester. On the CPU side, material set and assembly process used for electronic package of server CPUs often causes the lid of the package to have a warpage value in the range of 200-400 um. On the cooling solution side, its manufacturing processes, due to cost constraints, lead to non-flat surfaces. As a result, the two contacting surfaces where TIM2 lies are not flat and differ from test conditions used in TIM2 tester.In this paper, a model based on Hertzian contact, is presented to estimate TIM2’s effective thermal resistance in real usage. This model allows characterization data to be applied in situations where contact conditions differ from that used in the laboratory. Incidentally, it is found that impedance vs bond-line-thickness (BLT) characterization data is more appropriate than impedance vs pressure characterization data. The experimental data supporting the model are presented as well. Finally, it is pointed out that a more accurate estimate of TIM2 thermal resistance under usage condition can be obtained by using "in-situ" warpage/flatness of the contacting surfaces. How this in-situ warpage/flatness can be estimated via simulation and/or experimental methods is also shared.