The Sort-Assemble-Blend Routing Problem and Its Application to Semiconductors

Recent technological advances in the semiconductor industry have rapidly changed how products are designed and manufactured. Large “monolithic” designs, which integrate multiple functionalities in a single printed die, are being replaced with multidie products. To ensure the success of this transformation, it is critical to sort materials into categories and strategically build assemblies with maximum compatibility in mind. This enables manufacturing to be optimized to fulfill specific requirements for final products. However, it also means that more individual components must be managed, and decisions on how to assemble them must be made holistically. This creates a complex, nonlinear problem that combines qualities of both selective assembly and assembly routing problems. The result is named the sort-assemble-blend routing problem (SABR-P). Its solution approach is formulated using techniques including genetic algorithms, Monte Carlo simulations, linear programming, and machine learning. This framework has been successfully applied in practice, saving Intel Corporation more than $500 million thus far. It is the new standard for analyzing and optimizing Intel’s multidie products for the foreseeable future.