An important point to note in 10gR2 is that you
have two new db_file_multiblock_read_count parameters:

On my system:
  _db_file_exec_read_count               95
  _db_file_optimizer_read_count            8

leading to:
  db_file_multiblock_read_count           95

I haven't tested this yet (still doing lots with an explicit
db_file_multiblock_count, which forces both the others to
match the set value).

Working hypothesis: Oracle has finally separated the two functions
of the db_file_Mulitblock_read_count: one as the value to be used
by the optimizer (in the absence of an observed MBRC) for the arithmtic,
and one to be attempted at run time when executing the query). The MBRC
you finally gather will be probably be based on the actual run-time
activity, and will then over-ride the _db_file_optimizer_read_count
next time around the optimization cycle.

Regards

Jonathan Lewis


--
http://www.freelists.org/webpage/oracle-l