This paper presents a comparison of computational time required to build the admittance matrix of five test systems (ranging from 200 to 70,000 nodes) considering two formation approaches: element-by-element and matrix approach. The algorithms have been implemented in MATLAB TM and tested in four multi-core platforms. Implementations include sparse and dense matrix representation and parallel/non-parallel computing. Results show the matrix approach considering sparse representation and parallel computing is the best approach in computing time.