We use hydrodynamic cosmological simulations to study damped Lya (DLA) and Lyman limit (LL) absorption at redshifts z \ 2È4 in Ðve variants of the cold dark matter scenario : COBE-normalized (CCDM), cluster-normalized (SCDM), and tilted (n \ 0.8) models, as well as open (OCDM) and ) m \ 1 Ñat (LCDM) models. Our standard simulations resolve the formation of dense concentrations ) m \ 0.4 of neutral gas in halos with circular velocity km s~1 for and 90 km s~1 for v c º v c,res B 140 ) m \ 1 at z \ 2 ; an additional LCDM simulation resolves halos down to km s~1 at z \ 3. ) m \ 0.4, v c,res B 50 We Ðnd a clear relation between H I column density and projected distance to the center of the nearest galaxy, with DLA absorption usually conÐned to galactocentric radii less than 10È15 kpc and LL absorption arising out to projected separations of 30 kpc or more. If we consider only absorption in the halos resolved by our standard simulations, then all Ðve models fall short of reproducing the observed abundance of DLA and LL systems at these redshifts. To estimate the absorption from lower mass halos, we Ðt a power law to the relation between absorption area a and halo circular velocity in our v c simulations and extrapolate using the Jenkins et al. halo mass function ; we do not apply this method to the TCDM model because it has too few halos at the level resolved by our simulation. In the two LCDM simulations, for which DLA results agree well in the mass regime of overlap, the mean cross section for DLA absorption is much larger than the simple estimate based a B n(0.3R vir )2, a D n(0.1R vir )2 on collapse of the baryons to a centrifugally supported disk is the halo virial radius). The cross (R vir sections for LL absorption are with a dependence on numerical resolution at the D25% a B n(0.6R vir )2, level. Detailed examination provides further evidence of nonequilibrium e †ects on absorption cross section : for example, individual absorbers can be slightly smaller in more massive halos because gas sinks deeper into the potential wells, but more massive halos nonetheless have larger average cross sections because they are more likely to have multiple gas concentrations. Our extrapolation procedure implies that all four models are consistent with the observed abundance of DLA systems if the Ðtted extends to km s~1, and they may produce too much absorption if the relation continues a(v c ) v c B 50È80 to km s~1. Matching the observed abundance of LL systems requires absorption in halos down v c [ 40 to km s~1. Our results suggest that LL absorption is closely akin to DLA absorption, arising v c B 30È50 in less massive halos or at larger galactocentric radii but not caused by processes acting on a radically di †erent mass scale. Robust tests of cosmological models against the observed amount of high column density absorption will require simulations of representative volumes that resolve halos at the low-mass limit where they cease to harbor high column density absorbers, km s~1. 30 [ v c [ 60