This article shows how to implement the same DISTINCTCOUNT calculation in two alternative ways, measuring and comparing the performance in different reports. You will see that while DISTINCTCOUNT can be implemented using SUMX / DISTINCT, the DISTINCTCOUNT version is usually better. That is, unless the density of the reports is high and the calculation apply filters to the measures that do not correspond to the grouping granularity of the visualization – as is always the case using time intelligence functions. There are cases where SUMX / DISTINCT can offer better performance, but you have to clarify whether optimizing one report might slow down many others. Measuring performance using DAX Studio is the only way to know what to expect for your model and reports.
Read on for the full test.