Further more we use a drastically simplified geographical setup. Our model world ("the Waterplanet") has two polar islands and a long narrow continent connecting them. At the southern end of the ocean basin there is a gap in the continent, to allow for the effects Drake Passage (see this figure here for relevant geography) to be included. The 5 km deep ocean as random distribution of 2km high sea mounts on the bottom, and especially there are sea mounts covering the latitudinal extent of "the Gap".
With this very simple setup we can reproduce much of the observed features. Even if the model is run using only temperature as the active tracer, the oceanic temperature field obtained is in quite reasonable, provided that "the Gap" is open. If we run the model with the Gap closed we get very different results, with hemispherically symmetric ocean circulation.
Another related question is what effects the basic thermal structure has on climate. In our open Gap experiments there is considerable heat transport from the southern hemisphere to the northern hemisphere, which leads to a pole-to-pole temperature difference, sometimes of few degrees C. This asymmetry in hemispheric temperature structures is a direct consequence of the asymmetry imposed on the model domain through the inclusion of the Gap. Processes in the Gap (such as wind strength), can influence the pole-to-pole temperature gradient. The net result is that climate variability in the model northern hemisphere is to some extent remote controlled from the southern hemisphere.
For a postscript version of a recent poster about this (from the AGU Ocean Sciences Meeting, Jan 1999), follow this link here. An article on much of the same material has been accepted for publications in the Journal of Quaternary Science. Unfortunately I do not have a download-able version of it. Contact me if you want pre-prints.