Thermomechanical simulations are the core of many industrial applications such as metal additive manufacturing or nuclear fuel modeling, where the thermal gradients induce stresses and mechanical deformation affects temperature fields. Obtaining accurate coupled solutions is essential for process optimization and component safety. Coupled multiphysics problems can be solved using monolithic methods, i.e. solving all equations simultaneously through a given solver, or using partitioned methods that iteratively call monophysics solvers until convergence. While monolithic approaches offer enhanced accuracy, partitioned approaches are prefered here thanks to their greater flexibility, improving memory management and efficient solvers reuse. In partitioned thermomechanical coupling, the influence of mechanical displacement on thermal fields is often neglected or handled by solving the thermal model on deformed geometry. However, mesh deformation introduces limitations including additional computing and memory costs for mesh updates and field projections. An alternative approach exists in the literature but remains rarely implemented in industrial codes: an effective heat flux (sometimes called Piola-Kirchhoff or nominal heat flux) density and thermal conductivity are determined analytically to obtain equivalent temperature solutions on the initial configuration. This work investigates the effective conductivity approach for linear static thermoelastic problems with a Gauss-Seidel partitioned coupling strategy. Numerical equivalence between the deformed and initial geometry approaches is verified on a thermally loaded beam and a three-dimensional nuclear fuel pellet geometry. Both methods converge in identical number of fixed-point iterations with relative difference at machine precision. Computational analysis reveals comparable run times as anisotropic effective conductivity computational costs is offset by faster thermal resolution on undeformed mesh. Strong scalability tests on a cluster composed of AMD EPYC Milan CPUs obtained with MFEM HPC library (https://mfem.org/) demonstrate near-ideal speedup on to several thousand MPI processes on meshes with hundreds of millions of degrees of freedom. This demonstrates that effective conductivity provides a robust, equivalent alternative to mesh deformation without computational overhead, suitable for large-scale HPC simulations while preserving mesh quality.