Abstract. The Snowball Earth bifurcation, or runaway ice-albedo feedback, is defined for particular boundary conditions by a critical CO2 and a critical sea-ice cover (SI), both of which are essential for evaluating hypotheses related to Neoproterozoic glaciations. Previous work has shown that the Snowball Earth bifurcation, denoted as (CO2, SI)*, differs greatly among climate models. Here, we study the effect of bare sea-ice albedo, sea-ice dynamics and ocean heat transport on (CO2, SI)* in the atmosphere–ocean general circulation model ECHAM5/MPI-OM with Marinoan (~ 635 Ma) continents and solar insolation (94% of modern). In its standard setup, ECHAM5/MPI-OM initiates a~Snowball Earth much more easily than other climate models at (CO2, SI)* ≈ (500 ppm, 55%). Replacing the model's standard bare sea-ice albedo of 0.75 by a much lower value of 0.45, we find (CO2, SI)* ≈ (204 ppm, 70%). This is consistent with previous work and results from net evaporation and local melting near the sea-ice margin. When we additionally disable sea-ice dynamics, we find that the Snowball Earth bifurcation can be pushed even closer to the equator and occurs at a hundred times lower CO2: (CO2, SI)* ≈ (2 ppm, 85%). Therefore, the simulation of sea-ice dynamics in ECHAM5/MPI-OM is a dominant determinant of its high critical CO2 for Snowball initiation relative to other models. Ocean heat transport has no effect on the critical sea-ice cover and only slightly decreases the critical CO2. For disabled sea-ice dynamics, the state with 85% sea-ice cover is stabilized by the Jormungand mechanism and shares characteristics with the Jormungand climate states. However, there is no indication of the Jormungand bifurcation and hysteresis in ECHAM5/MPI-OM. The state with 85% sea-ice cover therefore is a soft Snowball state rather than a true Jormungand state. Overall, our results demonstrate that differences in sea-ice dynamics schemes can be at least as important as differences in sea-ice albedo for causing the spread in climate models' estimates of the Snowball Earth bifurcation. A detailed understanding of Snowball Earth initiation therefore requires future research on sea-ice dynamics to determine which model's simulation is most realistic.