We present a new method for whole-core Monte Carlo calculation using space domain decomposition to alleviate the excessive memory requirement due to massive tallies. The proposed method is called the fission and surface source (FSS) iteration method; it is based on banking both the fission and surface sources for the next iteration to provide exact boundary conditions for nonoverlapping local problems. To accelerate source convergence during inactive iterations, the p-CMFD (partial current–based coarse-mesh finite difference) method is applied to adjust the weights of the fission and surface sources. While domain-based parallelization is easily implemented using the proposed FSS iteration method, the computing times for the local problems will be different, depending on specific local problems, which may cause idle times of the processors to wait for the results from other local problems. To reduce the idle times, we apply a source-splitting scheme to the FSS iteration method to level the expected numbers of the sources of local problems. The performance of the FSS iteration method is tested on two-dimensional, continuous-energy reactor problems, with encouraging results.