PowrPivot Performance comparison


  • Hi There,

    I try to get advice on which data source will performance better for end user using Excel PowerPivot. I have SQL server data source in dimensional model (Fact table, dimensions) , I also have traditional SSAS cube developed based on my dimensional database, I also have new SQL 2012 Tabular model developed.

    Excel PowerPivot will be used for users, obviously I can connect all three data sources through powerpivot, but which one performance better in terms of retrieve time, memory usage, filter?



    Wednesday, June 19, 2013 7:00 PM


All replies

  • I would use SQL Server. Probably Tabular is faster, but reading one table at a time from SQL Server you shouldn't see a big difference. Moreover, with the current connection technology (based on SSAS ADODB driver) you get column names that are very bad, whereas in SQL Server you can provide to the user nice views with good names.

    Thursday, June 20, 2013 4:42 PM
  • Marco,

    Thanks for insight. Only thing I am not sure is that.

    If we import from SQL Server, does client pc need lots of memory since whole data set is at client, does this impact performance if data is too big.

    If we import from SSAS, I am guessing maybe ssas server will take memory hit, not client pc, same with tablular model.

    Lots of our users only have 32 bit machie with 2 or 3 gb of memory.

    Any more advice?


    Friday, June 21, 2013 3:59 PM
  • whenever you Import data into PowerPivot you hit the memory of the Client - this is regardless of the source

    PowerPivot always stores the whole dataset in Memory

    - -

    Monday, June 24, 2013 1:05 PM
  • Hi All,

    Are there any performance  /  technical / other benefits to using SQL as a data source vs CSV, Txt or linked files.


    Thursday, March 05, 2015 8:01 AM
  • well, SQL has predefined/fixed datatypes and structures which CSV does not have
    it is always more safe to import data from SQL than from CSV

    I would also expect performance to be better from SQL

    so long story short, go with SQL if you can :)


    Gerhard Brueckl
    blogging @
    working @

    Friday, March 13, 2015 1:55 PM