When would be the result of Cube & T-SQL different?

It depends on what kind of data are you looking for. Cubes give you pre-aggregated results whereas with standard T_SQL queries you have to aggregate them on the fly which puts additional load on hardware. So, by using cubes on large sets of data, with good understanding and definitions of measures and dimensions you can save a lot of time, both yours and servers's.

To answer your question, result data set should not be different (because it uses the same data source/Data Warehouse). It should be the same, but already aggregated, depending on many "variables": hierarchies, dimensions, measures... Slice, Dice, Drill-through...