Known good die, die replacement (rework), and their influences on multichip module costs

A systematic study of the known good die (KGD) percentage on multichip module yield has been made. In this study, repair is used to improve overall module yield, and cost models for both nonrepair and repair of defective die are developed. Results indicate that in both repair and nonrepair scenarios, minimums exist in the cost versus KGD percentage curves. These minimums suggest optimum KGDs percentages to minimize both chip and module cost. Repair is extremely important in cases of low repair cost or high die cost. Overall cost improvement can be achieved by paying for increased KGD percentage in cases of high repair (or module cost) or low die cost. The first repair is shown to be the most significant in improving overall module yield