For commercial grade semis then you can expect an operating temperature of the silicon die of 85C, and there was a general rule of thumb when I was a lad that you halved the MTBF for every rise of 10C above this. Note, I'm talking die temperature and that is not the same as the package temperature (which will likely be less, depending on environment). Of course what should be measured is the temperature increase - if your ambient air temp is 39C then a die temperature of 40C is less of a concern than an ambient of 10C.

BUT, the question you should be asking is more why the micro controller is getting warm. Actually, 40C isn't particularly hot, though I cannot recall ever fingering my UC300eth whilst in operation - I don't think it gets particularly warm (and I'm not about to strip it down to check). What you seem to be indicating is that the temperature of yours rises quickly and you expect it to continue to rise if not de-powered. That would concern me a lot. Is this a new device?, do you know if it is functioning correctly?, have you tried this with nothing connected to the input/outputs from the board?