Mixed-phase clouds play an important role in modulating the radiative energy budget at high latitude and can substantially impact regional and global climate change. However, large uncertainties exist in the representation of mixed-phase cloud properties in general circulation models (GCMs). In the recently released second version of the U.S. Department of Energy (DOE) Energy Exascale Earth System Model (E3SM) atmosphere model (EAMv2), new physics parameterizations are incorporated and new tuning parameters are applied to the convection and cloud microphysical processes during the model development from EAMv1. The goal of this study is to evaluate the simulation of mixed-phase cloud properties over high latitude regions in the northern and southern hemispheres, and understand the impact of new physics features on simulated clouds from EAMv1 to EAMv2. The GCM-Oriented CALIPSO Cloud Product (CALIPSO-GOCCP) product and ground-based remote sensing measurements from the Atmospheric Radiation and Measurement (ARM) program are used to evaluate model simulated cloud properties. CALIPSO satellite simulator and a novel mixed-phase sampling approach are used to address the cloud definition and data resolution difference between the model and observations for a consistent model and data comparison. Our results indicate that EAMv2 overestimates cloud liquid water in simulated Arctic mixed-phase clouds when comparing to both observational datasets. The evaluation with satellite data suggests that the underestimation of cloud ice in EAMv1 in the Arctic is largely improved. However, the integrated cloud ice water path remains underestimated in EAMv2 when comparing modeled clouds to the combined radar and lidar ground-based retrievals at the ARM North Slope of Alaska (NSA) site. This difference is probably due to the attenuation of lidar beam at cloud top, which limits the spaceborne lidar to detect cloud ice in the lower cloud layers. This difference also suggests the importance of combining different observations in the model evaluation to minimize the impact of limitations in individual dataset. Through model sensitivity experiments, we found that the new dCAPE-ULL convective trigger in EAMv2 substantially improves the cloud phase simulation over the Norwegian Sea in the Arctic and the Southern Ocean. This highlights the impact of convection parameterization on high latitude mixed-phase clouds in GCMs, which previously received little attention.