Some functions seem only use one core including initial stages of an analysis run. This is particularly noticeable for contact studies.
This can be frustrating for very large models and one should avoid animation in the post processor for large models.
Also some analysis types e.g. large deflection, fatigue (I think) only use one core.
There are some earlier threads on here may at least help to understand the limitations. These discuss the role of memory, disk and 64bit.
Can you explain more how it slows you down?