Yield estimates for the TESH multicomputer network

Reconfiguration and yield issues are considered for parallel computing systems based on a new interconnection network, "Tori connected mESHes (TESH)." The network is hierarchical - thus allowing exploitation of computation locality as well as systematic expansion up to a million processors, permits efficient VLSI/ULSI realization, and appears to be well suited for 3-D VLSI/ULSI implementation. Further, it enables efficient implementation of a large variety of computation intensive applications. In a very recent paper, we have reported the methodology for automatic reconfiguration of TESH. The goal of the present paper is to derive theoretical estimates of system yield as a function of the defect density and the various reconfiguration approaches. From these bounds, it is concluded that despite dramatic improvements in defect density in recent years, it is still necessary to provide redundancy and defect circumvention to achieve acceptable system-level yields. The analysis includes consideration of not only cell yield, but also switch and link yields. The latter two are particularly important in view of the emerging interest in interconnected nanotech devices for which the link yields currently are quite low.