Ten years ago, the conventional wisdom as cited by Rex Rice and others was that interconnect wiring on a silicon chip was much less expensive than interconnections on a ceramic hybrid, a printed circuit board, or cable interconnect. That led to a major emphasis on increasing the size and complexity of the silicon chip, with the other interconnect media left for the overflow or leftovers that could not be placed on the chip.

A major change of thinking was triggered by Knausenberger and Schaper of AT&T (1), with the realization that costs normalized per inch of wire length were nearly identical for all forms of interconnect. Literally an inch of interconnection circuit costs the same whether that circuit was on silicon or on ceramic, whether that circuit was on a printed circuit board or in cable.

If the only important criteria is the length of the interconnect, then a system or a board of the smallest size and area for a given circuit will have the shortest path lengths and the lowest cost. The dominant criteria is the area of the interconnection medium that carries the active silicon.