Hmmmm. I'm not sure, as I tried to indicate, whatever little I knew on the subject is shrouded in the mists of my mind, but I think that going via the ODBC layer to get to the named pipe would largely, if not completely, negate any (performance) benefits of doing so.

Note I wasn't suggesting named pipes were faster I was only saying you don't need specific support in DBI for them. I've no personal experience which would suggest they are faster or slower and would defer to MS on that without trying it myself.