Couple questions about partitioning and performance gain

Hello,

I try to maximize cube performance with partitioning. As a test, I have created a very thight partition for one customer only and compare the cube performance between this customer and other customers in larger partitions.
I am a little confused because I do not see a difference.
I also note that when I include calculated measures in the result set, it really knock-down the performance. example, this calculated measure:

iif([Margin Type].CurrentMember.name = "Historical",
iif(isempty([Measures].[AMT]) or [Measures].[AMT] =0, null,
([Measures].[AMT] - [Measures].[Margin1 Cost])/[Measures].[AMT])
, iif(isempty([Measures].[AMT]) or [Measures].[AMT] =0, null,
([Measures].[AMT] - [Measures].[Margin1 Cost Inception])/[Measures].[AMT]))

Really slow-down the cube even for the customer that has a dedicated partition.
The expected result is a matrix of 10,000 rows for 4-5 attributes and 4 columns of measures plus 4 calcuted measures times 8 quarters. unfortunately, I need that much data.

I tried to set the non-empty behavior of this calculated measure to be both types of cost and removed the isempty or 0 test but did not see any change.

I also did both a 30% aggregation plus the user queries optimization.

The fact table is 6M rows, 2 of the dim tables are 300K rows. I run SQL2005 enterprise RTM (no sp1 yet, I did not test it). I use Excel 2003 pivot as front-end.

So is there any benefit in creating lot of small partitions to boost most likely users queries?

I am learning, reading and trying but I lack the practical experience.

Any insights?

Thanks

Philippe

Does this perform any faster?

Create Member Measures.yourcalcmeasure = 1 - [Measures].[Margin1 Cost Inception] / [Measures].[AMT];

(Measures.yourcalcmeasure, [Margin Type].Historical) = 1 - [Measures].[Margin1 Cost]/[Measures].[AMT];

(and try to deal with "divide by zero" errors in the client application)

Thank you