This study investigates the reliability of SnPb and SnAg solder joints in semiconductor packages subjected to thermal cycling. More specifically, solder joint crack growth and life are experimentally measured, and FEM models are run to explain the test results. Ultimately a life-prediction model is proposed for both SnPb and SnAg solder joint packages. Motorola 357-plastic ball grid array packages on printed wiring boards were thermal cycled with the following test parameters: SnPb and SnAg solders, three post-process conditions (aged, aircooled and quenched), four package layouts on the printed circuit boards (singledense, single-sparse, double-alternating, and double-dense), three accelerated thermal cycling protocols (0°C to 100°C, -40°C to 125°C, and -55°C to 125°C), and tests run at Motorola and the University of Texas. At predetermined thermal cycles, packages were removed from the environmental chambers, dyepenetrated, packages removed to expose the solder joints, and optical images taken. Images were processed to measure crack area, shape, orientation and length to show crack growth. Selected joints were sectioned and polished to investigate microstructure and failure modes. Selected boards were connected to an ANATECH event detector to monitor life from joint failures. FEM crack initiation and propagation models were developed to better understand failure mechanisms. Major experimental results are: 1) SnPb joints have about 50% faster crack growth rates than SnAg joints, subsequently SnPb joints have half the life of SnAg joints, 2) air-cooled and quenched packages had similar failure characteristics, but aged SnPb joints had lower life and aged SnAg joints had significantly longer life than the comparable nonaged joints, 3) double-dense package layout significantly decreased life (by 75%) over the other package layouts, which were similar to each other, 4) the test results at the two locations (UT and Motorola) were similar for SnPb solder joints, but significantly different for SnAg solder joints, and 5) the largest cracks occurred at the corners of joints just under the die edge. Major FEM simulation results are: 1) the crack initiation life of SnAg joints is approximately 100% longer than SnPb joints, 2) shear load is a major cause of crack growth, but the contribution of tensile load increases as the cracks grow, 3) primary cracks at the board interface appear to reduce the propagation rate of the primary crack on the package interface, 4) secondary cracks are suppressed when compressive stresses prevent voids from nucleating, 5) the double-dense configuration shows no PWB warping due to symmetry, and its stresses are larger than for the other package layouts, and (6) the stresses and strains for single-dense, single-sparse, and double-alternating package layouts are similar because the stresses/strains are dominated by local effects due to the CTE mismatch between the die and board. Based upon the experimental results and FEM simulations, a lifeprediction model based upon a severity metric was proposed. The metric estimates damage to the solder joints and links material properties and parameters associated with package layout and thermal test conditions to the time-dependent creep, time-independent plastic deformation, and a time-dependent and geometric effective stress of the solder. The severity metric predicted life very well for most of the data tested and was more accurate than the industry-standard life-prediction models for SnPb solder joints.