Estimation of steady-state temperature field in Multichip Modules using deep convolutional neural network

•A model based on deep convolutional neural network is developed to predict the steady-state temperature field in Multichip Modules.•An end-to-end mapping is reconstructed from SDF presented chip placement and power to the temperature field of Multichip Modules.•Fast and accurate prediction with random chip placement is achieved, even under the conditions not included in model training.•A valuable tool is expected for the rapid thermal design of Multichip Modulus.