The effects of different kinds and amounts of fillers, e.g., carbon black N220 and silica-silane, on molecular structure and dynamics of styrene-butadiene rubber (SBR) and the effects of vulcanization on filled SBR systems are studied by Hahn echo and rotating-frame longitudinal relaxation T-1p NMR techniques as well as mechanical measurements, with the emphasis of quantitative comparison between microscopic and macroscopic results. The calculation based on a theoretic model for transverse relaxation in elastomeric networks reveals that filler aggregates or clusters behave like additional cross-linking points and thus lead to a decrease in the molecular mass between cross-linking points M-c, an increase of intercross-link chains, or accordingly a decrease of dangling and free chains, ultimately to the increasing cross-link modulus G(c), to the extent that depends on the type and, in particular, content of fillers. Compared with SBR filled with silica-silane, SBR filled with carbon black N220 shows more network chains due to its stronger binding ability or adsorbability but results in less cross-link density due to its higher inhomogeneity. The filler dependence of segmental and overall chain mobility is definitely different for both SBR systems, cross-linked and un-cross-linked, which is attributed to the restriction of filler on chemical cross-linking reaction occurred in a SBR matrix. A recent concept of hydrodynamic reinforcement by filler clusters and stress-induced cluster breakdown is combined with a tube model of rubber elasticity for modeling the typical stress softening effect of filled rubber for uniaxial stress-strain measurements. The evaluated material parameters deliver information about cross-link densities and filler specific polymer-filler couplings. The resulting densities V-h of network junctions are fairly consistent with NMR-derived ones within experimental errors; the differences can be used to estimate the content of different types of cross-links and chain entanglements. Finally, the related observations have been schematically represented using different physical models.