Originally posted by grjohnson
No, what I concluded running 9i, is that when using different MFBR figures in non_default tablespaces the FTS's scattered read will always (from my testing) perform the same number of "reads" that the DEFAULT tablespaces use.

Therefore, in this case, as stated, there isn't an advantage to large block size from a READ performance point of view in 9i using larger sized NON-DAFUALT tablespaces.
That's the correct conclusion with respect to default TS and DMRC, but consider what happens if you "up" the multiblock read count -- then you will start leveraging the larger non-default block size.

For a db with system TS of 8k blocks and a DMRC of 32, operations on a TS with 32k blocks are disadvantaged because they are effectively operating with a DMRC of 8. If much of your io is coming from the 32k TS you would do better to raise your DMRC -- say to 64.