The application of Ball Grid Array (BGA) technology in electronic packaging on high I/O plastic and ceramic packages has grown significantly during the past few years. Although PBGA (plastic BGA) has several advantages over fine-pitch Quad Flat Pack (QFP) in terms of smaller package area, higher I/Os, lower switching noise, large pitch, higher assembly yield, and improved robustness in manufacturing process, potential package reliability problems can still occur, e.g., excessive solder joint deformation induced by substrate warpage, moisture ingression (popcorn effect), large variation in solder ball size, voiding as a result of flux entrapment and improper pad/solder mask design (Marrs and Olachea, 1994; Solberg, 1994; Freyman and Petrucci, 1995; Lau, 1995; Donlin, 1996; Lasky et al., 1996; Munroe et al., 1996). Regardless of its improved thermal fatigue performance over the past few years through an extensive amount of research, the BGA solder joint may still pose a reliability issue under harsh environment, e.g., automotive underhood, larger package size, or higher temperature and temperature gradient due to increase in power dissipation of the package. Numerous studies in BGA solder joint deformation and reliability under thermal and mechanical loadings can be found in the literature, e.g., Borgesen et al. (1993), Choi et al. (1993), Guo et al. (1993), Ju et al. (1994), Lau et al. (1994) Lau (1995), and Heinrich et al. (1995). Also, reliability prediction models have been developed by, e.g., Darveaux et al. (1995) and Darveaux (1996). The present study focuses on the application of a detailed nonlinear finite element analysis (FEA) to studying the thermal cyclic response of solder joints in two particular BGA packages, full-matrix and perimeter. Both time-independent plasticity and time-dependent effect, i.e., creep and relaxation, are considered in the constitutive equations of solder joint to evaluate the discrepancy in the results of life prediction. The critical solder joint is identified, and the locations that are most susceptible to fatigue failure in the critical joint are discussed. Some limitations in computation and reliability prediction are also discussed.