How does version 6 differ from version 1?
The set of fields is the same, the order differs: in version 1 the timestamp starts with its low bits, in version 6 with its high bits. Because of that version 6 identifiers sort by time as plain strings, and version 1 identifiers do not.
Why does sorting by time matter?
It decides how new rows land in a database index. Ordered keys are appended to one edge of the tree, while random ones scatter across it, which forces pages to be loaded and split more often. On large tables the difference shows up in insert speed.
When should I pick version 7 over version 6?
Almost always, if the system is new. Version 6 is meant as a replacement for version 1 where such identifiers have already piled up: it keeps their fields, including the node address. Version 7 is simpler, stores ordinary Unix time and does not disclose the node.