Millivolt switches will support better energy-reliability tradeoffs

Summary form only given. Millivolt switches will not only improve energy efficiency, but will enable a new capability to manage the energy-reliability tradeoff. By effectively utilizing this system-level capability, it may be possible to obtain one or two additional generations of scaling beyond current projections. Millivolt switches will enable further energy scaling, a process that is expected to continue until the technology encounters thermal noise errors. If thermal noise errors can be accommodated at higher levels through a new form of error correction, it may be possible to scale about 3× lower in system energy than is currently projected. A general solution to errors would also address long standing problems with Cosmic Ray strikes, weak and aging parts, some cyber security vulnerabilities, etc.In conclusion, millivolt switches may be able to convert computer reliability from an annoying problem that can be “swept under the rug” into a key enabler for additional generations of scaling. Since the introduction of the integrated circuit, reliability of parts has been high enough that it could be dealt with through ad hoc techniques (like error correction for memory) and inefficient recovery methods (like checkpoint-restart). As the largest computers continue to grow to Exascale, the scalability of the existing methods is being tested and coming up short. While millivolt switches are not currently available, it can be projected theoretically that realizing their maximum potential will require a more systematic solution to reliability issues to address thermal noise. However, the argument just presented has transformed reliability from a problem into an enabler for increased power efficiency. This transformation is of more than scientific interest, as funding tends to be more accessible to projects that realize upside potential than projects that fix problems.