Since you'll always make a call from an EntityManager, the question is how does that EntityManager know or find an EntityServer to handle its queries, saves, or other calls. Generally in server-side code (like an interceptor or login manager) you'll be provided with an EntityManager - that EntityManager is always directly connected to the EntityServer, basically as a 2-tier application - in other words there's no network or out-of-process communication occurring.
If you "new" an EntityManager within server-side code and do not use the copy constructor, then that EntityManager will look at configuration settings by default to determine whether it should act in a 2-tier or n-tier manner - by default it will normally think it's using a local EntityServer and will therefore work like a 2-tier application. Prior to the 6.1.7 release there were (somewhat rare) use cases where a new EntityServer instance could be created, but generally the EntityManager will talk directly to the current EntityServer.
So, when you use an EntityManager from server-side code, unless you've specifically told it not to, it will not use a network connection and will do in-process calls to the current EntityServer instance.
If you think network calls are being made from a server-side EntityManager, check the debug log. If the EM is trying to use WCF to talk to an EntityServer the log will include messages about WCF proxies created for certain addresses. If so, then check your config file settings.