We present a detailed investigation of local dynamics of linear and cyclic poly(dimethylsiloxane) (PDMS) covering a wide range of molar masses. To aid interpretation of the experimental data, QENS measurements in the time scale from 2 to 200 ps and at Q = 0.3 to 1.8 Å–1 are complemented by theoretical calculations. These make use of a methodology developed by us elsewhere applicable to both simple chain models and real chains and applied here, for the first time, to cyclic PDMS. Analysis of the incoherent dynamic structure factor at T < Tm shows that the rotational motion of the methyl groups is unaffected by polymer topology. At higher temperatures, the QENS data are described by a model that consists of two dynamic contributions: methyl group rotation and segmental motion, the latter described by a stretched exponential function. Relaxation times of both linear and cyclic PDMS increase with increasing molar mass. Several features predicted by theory are also reproduced by the experimental data. We show, unambiguously, that rings have higher relaxation times for the segmental motion compared to linear chains of the same number of monomer units. Theoretical calculations support the idea that such slowing down of local dynamics is due to the topological constraint imposed by the ring closure, a constraint which becomes negligible for very large molar masses. Our calculations suggest that due to its albeit small conformational rigidity, cyclic PDMS undergoes an additional constraint which further increases the relaxation time, producing a shallow maximum for N ≈ 50 repeat units. A similar feature is also observed in the experimental QENS data. Values of activation energy, Ea, are derived from analysis of the temperature dependence of the quasi-elastic broadening and are found to be in agreement with viscosity measurements reported in the literature. Although the pronounced molar mass dependence of Ea for linear PDMS is certainly linked to the presence of mobile chain ends, for the cyclic polymers the behavior appears to be more complex than anticipated.