We examine peak-to-trough gravity anomaly amplitudes over seamounts, comparing gravity derived from satellite altimetry with gravity measured by ships. We also compare amplitudes from linear and higher-order (non-linear) computations, and conclude that two-term forward modeling of seamounts should be adequate. Altimetric amplitudes are within 90% of ship amplitudes at seamounts greater than 14 km in characteristic radius. At smaller seamounts, altimetric amplitude is reduced, probably due to smoothing filters used to process the altimeter data. Perhaps fortuitously, this amplitude loss nearly matches the amplitude under-estimate that occurs if one assumes the forward model should contain only a linear term. Thus bathymetric prediction by linear filtering may do a fairly good job of resolving seamounts much smaller than 14 km in characteristic radius.