Question about AAS and Premium Capacity CPU usage

Hi everyone,

I'm currently experiencing huge performance decrease on our model. Unfortunatly no new measure or table has been changed since last week and the performance issue is sudden. 

I think this has to do with how our model has been built.

Here's the template :

For some context, Dataset A is our main dataset for sales. It's in Direct Query with an AAS Cube which hold 99% of the data, just a few Sharepoint dimension are added to it. 

The dataset B allows for the consolidation of multiple business datasets to obtain a unified view of certain KPI in a single report.

The performance drop is linked to our Premium Capacity usage, we're at +100% on the busy hours, while our AAS Cube only top at 50%.

My questions are the following :

Imagine two measure :

Sales A = SUM(Sales) which is exclusively from AAS

Sales AShare = calculate(Sales A,Sharepoint condition)

First, i guess Sales A use AAS CPU to calculate the result, am i wrong ? even if the measure is created in the PBix and not AAS?

Then does, Sales AShare , which use a condition from an imported table, is still using AAS CPU or does it use the PBi Premium Capacity CPU?   

And last question, if i use Sales A but from a report from Dataset B, will it still use AAS CPU or does it use the PBi Capacity CPU?

I don't know if the questions are clear enough for my issue,

Thank you