Monitoring Cube Performance and Usage - Expert Cube Development with SSAS Multidimensional Models - page 318

In other words, on an 8-core server we will see a constant consumption of one-eighth
of our available CPU, that's to say 12 to 13 percent (you see a slightly higher percentage
in the following screenshot because of the time spent by other processes).
In such cases, there are two ways we can try to optimize query response time.
On one hand, we can try to change our cube design or our MDX query to enable
greater parallelism during query execution; on the other hand, we can try to reduce
the time needed for a sequential operation, for example, by reducing the size of
an intermediate fact table used in a many-to-many dimension relationship, or by
rewriting MDX to use bulk mode instead of cell-by-cell mode (as discussed in
Chapter 8, Query Performance Tuning ).
In general, Analysis Services can consume a lot of CPU resources during processing,
and depending on the conditions, also while running queries. We need to bear
this in mind when specifying servers for Analysis Services to run on, or when
deciding if Analysis Services should be installed on the same server as other
CPU-intensive applications.