Tropical Cyclones (TCs) are typically not properly represented in global climate models, as GCMs are routinely deployed at low resolution, in order to enable sufficiently long integrations and/or ensemble members. A new class of GCMs is emerging, capable of simulating TC-type vortices by retaining a horizontal resolution similar to that of operational NWP GCMs, while permitting multi-decadal integrations. In the past, using a hierarchy of GCMs based on the HadGEM1 model formulation, we have shown how the evolution of GCMs from standard (200-300km) to NWP-type (~50km) resolution significantly improves the simulation of TC activity, including location and frequency. The simulation of interannual variability in the Atlantic basin for the 1978-2004 period, is particularly skilful in this new class of GCM. We now consider the skill of coupled atmosphere-ocean GCMs (the latter eddy-permitting) over a centennial time scale, in simulating the location, frequency and nature of tropical cyclones (activity, structure, lifetime evolution). We show how slightly lower SST values simulated by the AOGCM slightly reduce the total TC count, albeit to a minor degree, and how interannual to decadal variability in oceanic conditions plays a large role in TC activity. For instance, the location of Pacific TC activity is greatly affected by ENSO variability. The analysis of centennial scale coupled atmosphere-ocean integrations alongside atmosphere-only integrations not only allows us to investigate the impact of climate variability on TC activity, but also the impact of TCs on the global climate system, for instance in terms of eddy activity and relative transports, which alter the tropical/extra-tropical distribution of tropospheric moisture. With regards to the maximum TC intensities at each model resolution, this relationship is largely unchanged by the use of observed (AGCM) or internally simulated (AOGCM) SSTs. However, there is a slight reduction in TC intensity with ocean coupling, which is likely due to the negative feedback on SSTs leading to a weakening of the storm, which is not permitted in the uncoupled model. We also investigate whether this response is related to the higher degree of consistency between surface warming and lower stratospheric cooling (affecting TC outflow) in the AOGCM, which has more degrees of freedom in adjusting its global radiative balance.