Front-end dielectric requirements for end-of-the roadmap CMOS and beyond are considered. For the former, we focus on the surface-roughness limited channel mobility and short channel device performance, and on how use of alternative dielectric gate stacks with perhaps larger surface roughness might affect short channel device performance. We find that surface-roughness limited channel mobility can be a misleading predictor of short channel device performance, and, in particular, that perhaps increased surface roughness associated with use of high-k dielectrics may be far less problematic than channel mobility measurements would suggest. Dielectric requirements for two beyond CMOS device concepts, the heterobarrier tunnel FET (HetTFET) and the even more exotic graphene bilayer psuedo-spin FET (BiSFET) are also considered. While these latter devices may or may not ultimately work as hoped, they serve to illustrate how the demands on front-end dielectrics could change radically with emerging technologies, and indeed, how the emergence of beyond CMOS devices could depend critically on advances in dielectric technology.