@Jorge Marin Isn't thermal dissipation related to efficiency losses and it should vary from part to part depending on the application? Maybe a good place to start would be the already existing design of a power transistor that one of the Stanford students had created. You can take a look at the layout and determine how much channel resistance and metalization resistance is present in the layout. You can then probably size your transistor to meet your requirements. From the pdk you know how much the rdson varies with temp, ( I am guessing about 25% to 30%) and the metalization R remains the same roughly with temp( please check model and or documentation for temp coefficient information). You can then determine how much is the thermal loss just because of Rdson variation. In addition you will have to include the package thermal resistance as Boris mentioned. I hope this helps. Not sure if I answered any concern of yours. Maybe I made it too simple. :)

@Boris Murmann @Kamala Hariharan thanks for yous replies These are very helpful ideas :)