As with all things technical, by the late 1980s and early 1990s, IC reliability had improved dramatically as manufacturing processes and quality controls evolved (Moore’s Law at work). Meanwhile, as noted above, the market for ICs shifted from a focus on military to computers and telecommunications and then more recently to a focus on consumer electronics. With the IC industry emphasis now on personal communications and consumer-driven electronics, Mil/Aero OEM customers were able to obtain more complex devices. Meanwhile, the IC industry experienced far more rapid product evolution and shorter life cycles. The end result, of course, was the displacement of older ICs by newer devices. In short, while ICs were advancing, they were also being obsoleted in growing numbers and at a faster pace than perhaps anticipated by anyone.
Seeing only the rapid advancement in performance and the concomitant price decline of these new “commercial-oriented” devices, the military community longed for the improved performance (and lower cost) of these ICs. These commercial devices offered more function/performance per square micron of silicon than the traditional mil-spec parts. Moreover, the prices of the new ICs were dramatically lower. The enhanced reliability of the new generation of ICs, derived from the much improved manufacturing processes, leads many to postulate that the use of COTS devices should be seriously considered. More performance and at a lower cost—a true win-win!
Of course, this weighed against the staid (tried and true) view that a certified mil spec part would offer the desired long term reliability.
Into this dynamic situation stepped the famous Admiral Perry who effectively argued for, and essentially mandated, that government-funded programs use COTS parts. The government’s rationale was lower cost with higher performance from the latest technologies and much improved reliability. So Program Offices began to call for COTS parts as the means to achieve Adm. Perry’s call for “cost containment” for the ever more costly military applications.
The move to COTS parts proved successful in many applications. Performance improved, costs dropped, and the latest technologies became mainstream. COTS parts were holding sway across the majority of procurements and saving money for the majority of programs. And where COTS were not feasible, the more costly, full mil-spec parts could rationally still be used.
But time inexorably marched on. As new military and aerospace programs were initiated, many OEMs were anxious to utilize COTS parts. Further, some of the original COTS-based programs were extended beyond the originally planned lifetime horizon. An unintended and unanticipated problem emerged. Many of the new COTS-based systems had originally been designed with the former mil-spec parts which were expected to be available for the long term. In actuality, the COTS parts have finite lifetimes that are substantially shorter than the traditional mil-spec parts. Many OEMs discovered that a COTS part which they designed in had become obsolete long before the program into which the parts were deployed reached its end of life (Figure 1).
Figure 1: The life cycle for an IC in a consumer application is much shorter than for an IC in a mil/aero design. Over the same period of time the IC in a consumer application is expected to be revised frequently and generate revenue with each revision, while the mil/aero design generates revenue based on the long-term, original IC application.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.