Merge geometry or tie constraint

[Staszkinson]

Hello,
What type of modelling in Abaqus do you prefer?
I mean when it is recommended to merge geometry from parts into single large part?
How often do you use tie constraint?
As far I know the effect is almost the same - we can tie together parts of the geometry - we can tie together deegres of freedom at nodes.

What is influence of using tie constraint/merging mesh on results of the analysis?
Let's assume, that in both cases the finite element mesh is the same.

Thanks.