I'll give that a try but my instinct tells me that is not the issue - it seems to 'read' the first 9000-ish rows quickly, then takes an inordinate length of time to do the commit, its like its reading the buffer, then inserting and committing one row at a time. I'm not doing anything different to the way I would normally do it in sql server 2k5 or 2k8r2.