Ten 3-D cloud-resolving model (CRM) simulations and four 3-D limited area model (LAM) simulations of an intense mesoscale convective system observed on 23–24 January 2006 during the Tropical Warm Pool–International Cloud Experiment (TWP-ICE) are compared with each other and with observations and retrievals from a scanning polarimetric radar, colocated UHF and VHF vertical profilers, and a Joss-Waldvogel disdrometer in an attempt to explain a low bias in simulated stratiform rainfall. Despite different forcing methodologies, similar precipitation microphysics errors appear in CRMs and LAMs with differences that depend on the details of the bulk microphysics scheme used. One-moment schemes produce too many small raindrops, which biases Doppler velocities low, but produces rainwater contents (RWCs) that are similar to observed. Two-moment rain schemes with a gamma shape parameter (μ) of 0 produce excessive size sorting, which leads to larger Doppler velocities than those produced in one-moment schemes but lower RWCs. Two-moment schemes also produce a convective median volume diameter distribution that is too broad relative to observations and, thus, may have issues balancing raindrop formation, collision-coalescence, and raindrop breakup. Assuming a μ of 2.5 rather than 0 for the raindrop size distribution improves one-moment scheme biases, and allowing μ to have values greater than 0 may improve excessive size sorting in two-moment schemes. Underpredicted stratiform rain rates are associated with underpredicted ice water contents at the melting level rather than excessive rain evaporation, in turn likely associated with convective detrainment that is too high in the troposphere and mesoscale circulations that are too weak. A limited domain size also prevents a large, well-developed stratiform region like the one observed from developing in CRMs, although LAMs also fail to produce such a region.