Marco Russo and Alberto Ferrari do some counting:
The VertiPaq engine is basically data type-independent. This means that it does not matter whether a column is a string, a floating point, or a date: because of the dictionary encoding happening inside VertiPaq, all these data types use around the same amount of memory and perform at nearly the same speed.
However, when mixing different data types in the same expression, DAX will likely need to perform conversions between data types. Some of these conversions are nearly free, whereas others require the intervention of the formula engine, with a related performance impact.
We have already written about possible errors occurring during data type conversion here: Understanding numeric data type conversions in DAX and here: Rounding errors with different data types in DAX. The issue with conversion errors is mostly due to the fact that the precisions of fixed decimals (also known as Currency) and decimals (also known as floating point) are different. This article starts with a focus on performance.
Read on to see what Marco and Alberto have for us this time.