Depending on the speed of your source database - it's very likely that analysis services will be faster. This is the primary reason most people use it.
1. When you have 100+ of measures and dimensions
2. To keep one standard source of data and calculations ( Many situations users create their own calculations and report different figures)
3. Data is pre-aggregated and faster than a SQL table. (Table extract is faster than cube)
1. You cannot use the complete capabilities of Tableau like create custom Group in dimensions, using dimension in if statements etc
2. Cannot create Tableau extract from a cube, so you cannot export packaged workbooks