This paper proposes a novel methodological framework for effectively measuring the production frontier performance (PFP) of macro-scale (regional or national) R&D activities themselves associated with two improved models: a non-radial data envelopment analysis (DEA) model and a nonradial Malmquist index. In particular, the framework can provide multidimensional information to benchmark various R&D efficiency indexes (i.e., technical efficiency, pure technical efficiency and scale efficiency) as well as the total factor R&D productivity change (determined by three components: “catch-up” of R&D efficiency, “frontier shift” of R&D technology as well as “exploitation” of R&D scale economics effect) at a comparable production frontier. It can be used to not only investigate the potential and sustainable capacity of innovation but also screen and finance R&D projects at the regional or national level. We have applied the framework to a province-level panel dataset on R&D activities of 30 selected Chinese provinces.