N1QL USE KEYS vs bucket.get

I was wondering if there is any info on the performance and resources required when looking up a doc by key vs using the bucket.get option. As much as i understand the N1QL query will use the Get function as well under the hood, buti am interested in knowing how large is the overhead and if it needs additional resources to do this to determine which route i will take to get docs where i know the key.

If you already know the key and don’t need any further processing or will able to process in the SDK go ahead and use bucket.get()

N1QL 2 hops (Data node ==> query node ====> client) vs bucke.get 1 hop