Thermal management of electronics is becoming the main challenge due to the continual increase in power density. In device-level electronics, besides the temperature of multiple heat sources (also known as junction temperature), the temperature of the enclosure (also called skin temperature) should also be controlled to protect the user experience. In this case, the generation of the device-level compact thermal model for fast predicting the skin temperature will not only help to improve both the thermal design efficiency and temperature control strategy. In this paper, dynamic compact thermal models of commercial laptops are first generated based on the convolution method. Under the assumption of linear time-invariant (LTI) systems, the skin temperature of laptops could be fast predicted once the step response of each heat source is calculated. Compared to the computational fluid dynamics (CFD) model generated in ANSYS Icepak, the maximum transient deviation of the convolution method is about 5%. A linear fitting model is further proposed to online predict the skin temperature. The results show that after training the scale factors, the linear fitting model could accurately predict the skin temperature after 60 seconds (with a maximum deviation of 7%). The aforementioned results indicate that the proposed fitting model has tremendous potential for future optimization of the skin-temperature-constrained control strategy.