Why Does Moving Data Between Partitions Take So Long ?

I have a table with millions of rows and about 70 columns that move through a number of states (11 possible states in all) from "New" via various states to "Processed" and eventually to "Archive" (there's a complicated state diagram that I won't bore you with)


Movement between states is based on a heap of business logic including the move to Archive (not just dates).


Different sorts of processing (querying and update both by users and overnight processing) are carried out on the data according to its state.


Maintaining the indexes for optimum performance across the board is a headache.


We have two problems in that we want better query performance and want to be able to easily switch out objects that are in the Archive state.


I had in mind partitioning the table (and its indexes) on state so that :
(a) Queries would be directed only at the appropriate partition (that is always use "where state=" as part of the query)
(b) The Archive partition could be swapped out of the table periodically


In my test setup 10 of 11 partitions are in [PRIMARY] but Archive is in a different filegroup.


Query performance is OK - execution plans look good.


However my update performance is now appalling when moving between any two states (10 times as long as on the unpartitioned table).


I understand that when you update a column which is used as a partition key it will cause the row to "move from one partition to another" as it says in another post.


Fine - because that's exactly what I want - logically.


I can also understand that moving from one filegroup (and hence the underlying file) to another must mean that the data has to physically move.


However is the data physically moving whenever you move between partitions or what's going on to cause such a degradation in performance ?