>>>We know the:
>>>A is indexed on column ID with index T_A_ID
>>
>>I'd suggest that you create an index on on the id column specifically.
>
>As I said there is an index on ID
>Statistics is updated, and indexes, with reindex.
>The server is 2 processor with HT and 4GB RAM.
>
>This is happen even with a table wich has ~18milion records and the query returns ~20000; this is not a selective query?
>
>The query plan with forced indexes is different than query wich use indexes by itself, and has lower performances.
>I tested with SQL2005 for 1 day and from what I see there are some changes with execution plans. Same query on same data wich is catastrofal on 2000 work very well on 2005 and is optimized - The work of VFP team at core of SQL are getting rezults.

I was asking if there was an index specific to the field being joined.

I recently had a problem with a very, very slow statement using a join. One of the tables had a compound key with one of the columns used in the join. Adding an index specific to the column took the query, literally, from minutes to sub-one second.