Version 3 of the Energy Exascale Earth System Model (E3SM) was released in 2024 and includes a state-of-the-art fully coupled sea ice model. The model solves sea ice mechanics on an unstructured 20-30km mesh shared with the ocean model with vastly improved coastal sea ice dynamics compared to previous E3SM versions. Icepack from the CICE-Consortium is used to simulate sea ice column physics and coupling with the atmosphere and ocean. Therein, SNICAR-AD radiative transfer is affected by the morphological aging of snow atop a sea ice thickness and floe size distribution with mushy-layer thermodynamics and melt ponds. We quantify skill and bias of five E3SM ensemble members from 1850 to 2025 using historical and SSP2-4.5 forcing against orbital measurements of sea ice freeboard, coverage, and motion. The late 20^th century model exhibits global and hemispheric extent, concentration and drift commensurate with observations at the March equinox but is negatively biased globally at the September equinox. On average, 8% (1%) of sea ice volume is lost per decade at the March equinox peak as compared to 25% (8%) at the September equinox in the northern (southern) hemispheres from 1980 to 2000. The March-September Equinox Asymmetry amplifies in the 21^st century even though E3SM’s global surface temperature closely tracks the HadCRUT5 analysis in all ensemble members. 5% (2%) of sea ice volume is lost around March equinox peak in contrast to a 28% (22%) of the September loss per decade from 2000 to 2020. Steady decline in simulated Antarctic sea ice in the 21^st century displays little ensemble variability in contrast to observations of interannual extent variability since 2010. Most global sea ice loss occurs in the Arctic, outpacing ICESat and ICESat-2 for which the contribution of snow to freeboard is disentangled using a satellite emulator. We attribute our modeled Equinox Asymmetry to coupling physics targeted for improvement in E3SM Version 4 to address the overarching point that sea ice bias is tied to atmospheric concentrations of CO₂ in our model.