ORM's have always kinda shoehorned objects into the RDBMS data model and the abstraction isn't always pretty.
With NoSQL, if you're just trying to store some cache or session objects and you need super high read/write time, NoSQL shines. If you have massive datasets, like "timeline statuses" and you need to search for the needle in the haystack, NoSQL (map reduce) shines.
If you're trying to analyze what region of the united states your product sells best to the 12-18 year old range, then RDBMS is usually the ticket.