Iceland is the type example of a ridge-centered hotspot. It is controversial whether the seismic anomaly beneath it originates in the lower mantle or the upper mantle. Some recent studies reported that the 660-km discontinuity beneath central Iceland is shallow relative to peripheral regions and this was interpreted as an effect of elevated temperature at that depth. We investigate topography of the major upper mantle discontinuities by separating the effects of the topography and volumetric velocity heterogeneity in P receiver functions from 55 seismograph stations. Our analysis demonstrates that a significant (at least 10-km) shallowing of the 660-km discontinuity is only possible in the case of improbably low seismic velocities in the mantle transition zone beneath central Iceland. If, as in previous studies, lateral velocity variations in the mantle transition zone are neglected, the data require a depressed rather than an uplifted 660-km discontinuity. For a reasonable S-wave velocity anomaly in the mantle transition zone (around − 3%) no topography on the 660-km discontinuity is required. This can be explained by the lack of temperature anomaly or an effect of two phase transitions with opposite Clapeyron slopes.