Physical limits on ground-motion parameters can be estimated from spontaneous-rupture earthquake models but are subject to uncertainties in model parameters. We investigate physical limits at Yucca Mountain, Nevada, and assess sensitivities due to uncertainties in fault geometry, off-fault rock strength, the seismogenic depth, fault zone structure, and undrained poroelastic response of the fluid pressure. For the extreme scenario of nearly complete stress drop on the Solitario Canyon fault, peak ground velocity (PGV) at a site near the fault is sensitive to deep fault geometry and cohesive strength of shallow geologic units, while it is relatively insensitive to fault zone structure, the seismogenic depth, and pore-pressure response. Taking previous estimates of Andrews et al. (2007) as a benchmark, a 10° reduction in dip (from 60° to 50°) of the Solitario Canyon fault at depth, combined with doubled cohesion of shallow units, can increase both horizontal and vertical PGVs by over 1 m/s, to values exceeding 5 m/s. In a lower stress-drop scenario (constrained by regional extremes of coseismic slip inferred from the paleoseismic record), PGV is most sensitive to fault geometry at depth, is only modestly affected by fault zone structure, and is insensitive to cohesion of shallow units and pore-pressure response. Effects of rock strength on spectral acceleration are significant only at short periods (i.e., less than 3 s). The dipping normal-fault models predict asymmetric inelastic strain distributions with respect to the fault plane, with more intense inelastic deformation on the hanging wall, though that asymmetry may be moderated by poroelastic effects.