Multidimensional computations were made of spark-ignited premixed-charge combustion in a pancake-combustion-chamber engine with a centrally located spark plug and in two pent-roof-chamber engines, one with a central spark plug and the other with dual lateral spark plugs. A global combustion submodel was used that accounts for laminar kinetics and turbulent mixing effects.The predictions were compared with available measurements in the pancake-chamber engine over a range of loads, speeds, and equivalence ratios. In all cases the computed and measured cylinder pressures agreed well in trends and magnitudes (within 8%) for the entire duration of combustion. Fair agreements were also obtained between predicted and measured values of wall heat flux and emission index of nitric oxide. In the pent-roof-chamber engines the predicted maximum cylinder pressures also agreed well with measurements (within 12%) in cases with MBT (Minimum spark advance for Best Torque) or advanced spark timing. However, the agreement deteriorated (to about 27%) at very retarded spark timing.Although these model predictions are sufficiently good for engineering applications, a sensitivity study was performed to understand the sources of discrepancies between the predictions and measurements. It was found that the discrepancies could be explained through uncertainties in the gas temperature, turbulence intensity and length scale existing in the chamber prior to combustion. This finding emphasizes the need for accurate measurements in order to improve estimates of the initial and boundary conditions and the turbulence-model constants.