An accurate and efficient thermal simulation approach based on proper orthogonal decomposition is applied to predict the temperature profile in an 18-core CPU. The power consumption for the CPU is generated using gem5 and McPAT for the thermal simulation in the training and validation of the approach. It has been demonstrated that the proposed approach can offer an accurate thermal prediction of the CPU with a reduction in degrees of freedom by more than 5 orders of magnitude and a speedup of 3 orders of magnitude, compared to the finite element method with identical thermal simulation settings.