TINTE is a well-established code for the pebble-bed high-temperature gas-cooled reactor (HTR), including the complicated nuclear module and thermal-hydraulic module, which has been validated by experiments and widely used in the transient behavior simulation. However, only an operator splitting scheme is employed in TINTE to couple the neutronics and thermal hydraulics, and some physical quantities are not consistent in time. As a result, the accuracy and stability are limited by the additional error term derived from the unconverged physical term. In this paper, a fully implicit coupling method was investigated in which the coupled nonlinear fields at each time step are converged using Picard iterations. A physics-based preconditioning is proposed in the work here to further improve the computational performance of the fully implicit coupling method. Seven test problems are implemented based on a practical engineering model, rather than a simple model, to evaluate the performance of the Picard method. The numerical results show that the fully implicit Picard iteration method is more accurate and more stable, which permits longer time steps and a reduction of the computational burden for solving the coupled field equations. The computational efficiency is further enhanced when the physics-based preconditioning is utilized.