Absolutely agree on the performance front. If you're starting out, generally SQL Server will run well out of the box, and if you upgrade hardware and reinstall on the new hardware it'll generally work well without having to do anything else. At some point you might have to start thinking about I/O and how best to arrange files but maybe not for a while, depending on what you're doing. (Just to be clear, let me restate: I'm talking starting out here - this is obviously NOT the approach you'd take when building or upgrading a large, complex application - I'm talking basic CRUDdy type apps here.)
Contrast this with some of the NoSQL platforms where immediately you're forced to think about topics such as sharding and clustering (I'm looking at you Couchbase) and the pain in the ass factor is much lower.
Key point: you don't need much of a clue about SQL Server administration to start using it, but you perhaps do with Couchbase, Elasticsearch, etc. From an ease of use point of view with NoSQL, I'd probably point to MongoDB as having the most approachable learning curve.
There's also no BS about having to keep everything in memory to keep performance acceptable (I'm looking at you, again, Couchbase - that's two strikes). The whole point of SQL Server and other RDBMSs is that they continue to work well when the amount of data you need to store vastly outweighs the amount of physical memory you have. Of course, if you can get everything into physical memory it will perform better but, even if you can't, it will still perform well.
Of course, if you're using SQL Server you do need to learn SQL (true for any RDBMS), which can be off-putting. It's absolutely not a beautiful language - in fact it's bloody aggravating when you write a lot of it - but it is extremely powerful and excellent for working with sets of data arranged in rows (i.e., tables). Nothing else really comes close.