Setting where calculated members in an OLAP cube should be processed

This post says:
"Because, for example, if your query return less then 1000 members by dimension in your query, all the data is retrieve to the client and the calculated members are calculated on the client side."

So actually the calculated members are by default calculated by the client. This explains why so much CPU is used on the client when there are a lot of calculated members.

To set the location for calculating the members to the server, you can use this parameter in the query string:
Execution Location=3