An update of a data row in SQL Server is sometimes the equivalent of a delete and insert operation. But, if you don't change the clustered index key, and you don't change the size of any variable length data (e.g. varchar columns), then the update can occur in place.
If the clustered index key does not change in the update, then no indexes need to be changed (unless affected by the update, of course), regardless of whether the update was in place or not.
If the update cannot be done in place, then the row may still be stored in its current page, if there is enough room (see Fill Factor).
Doing your own delete and add would be
way slower.
If you are making design decisions based on your (mis)conceptions of the way SQL Server handles its internals, or based on the way you know or think it is done in other products, I strongly suggest you do yourself a favor and get a copy of the book "Inside Microsoft SQL Server 2000", by Kalen Delaney. It's an excellent book and it discusses, sometimes in excruciating detail, the internals of SQL Server and how it works.
Jeff Mason
Custom Apps, Inc.
www.custom-apps.com