The present paper describes the implementation and data structure issues involved in the explicit partitioned time integration sub-cycling strategies for transient heat transfer in a data parallel environment on a massively parallel platform such as the CM-5. Approaches to handle explicit multi-time step strategies for thermal problems on massively parallel computing platforms have not been addressed to date. A partitioned strategy based on the data masking of elements during subcycling are investigated in this study. The explicit finite element formulations used in this partitioned strategy, for applicability to massively parallel computations utilize the flux based element representations and explicit solution algorithms with various groups of elements in a physical domain having different time steps. In principle, such issues serve to permit an optimal time step selection when stiff and flexible element(s) exist in a computational finite element mesh. Studies indicate that partitioned strategies involving masking indeed improves the execution time, however, they have the bottleneck of reduced FLOP RATE performance as some of the processors are idle during subcycling in a given thermal analysis computation.