Professor Excel

Comments 26

  1. AC

    Thank you for this – very useful article, particularly for somebody trying to figure out how to make using Excel on a Mac a better experience (looks like Bootcamp is the way to go)

  2. Charles Trudel

    In response to this,
    Therefore if you got an idea of why 3 threads are slightly faster, please let us know.

    Well, my guess would be that having three threads, it takes three cores for them, the fourth core is use by the operating system.

    Where as if we use 4 threads, the 4th core has a conflict with the system.

    Does that make sense?

    • Henrik Schiffner

      Hi Charles,
      Thanks for the feedback. Unfortunately, there is not much information about this issue available. That leaves me with guessing. Therefore, your idea could be correct. I would just like to add, that it highly depends on your workbook. As far as I know, Excel tries to divide the calculation performance by the calculation chains. Because of that, the performance could differ for different workbooks.
      Can I also encourage you to measure it for your own workbook and environment? The test workbook can be downloaded above (or slightly improved from this page: Please let me know what you find out 🙂

  3. Sam BC

    thumbs up for the article.
    it would be interesting too, if by any chance you guys will test out the excel performance difference between laptop vs desktop running similar specs (quad-core processor, 4gb ram, ssd) 🙂

  4. David

    Great article someone who lives and breathes spreadsheets. Thank you for the depth of the research.

  5. Galant Koh

    Thanks for this. Very interesting and useful. There was one aspect I was interested in but that wasn’t included though. I don’t know if you’ve investigated this before or since. Do you know if, all else being equal, more RAM or a faster CPU make more difference?

    For example, an Intel i5 with 4GB or 8GB RAM. vs an Intel i7 with 4GB or 8GB RAM?

    I’d be interested to see which PC components give greater speed for Excel.

    Thanks though. Good read!

  6. jure


    I have one problem with excel on my brand new laptop. It is not working very smooth, when I scroll faster there are grey area showing up for a moment, it looks like the data on the screen is refreshing to slow. I have UHD display, maybe this is a problem. The laptop is Thinkpad T570 i7 7600, UHD display, 16GB RAM DDR4, Invidia GeForce 940mx, 512SSD.. So it is laptop with good specs but excel is working worst than my 8 years old HP. Do you have any suggestions? Thank you for your help, Jure

  7. JMuler

    after several tests with different Excel versions this is my conclusions:

    -Windows Server 2016
    -Excel 2007 (32bits) Excel 2013 (32 bits) Excel 2016 (32bits) (64 bits discarded due to slow performance)
    -Excel Workbook with 54 sheets and some vba macros.


    Faster Excel with VBA code is Excel 2007.

    But the fastest combination I found is Windows Server 2003 (32bits) with Excel 2007 (32 bits).

    And … If you put the affinity to only one CPU, the time is 50% faster. INCREDIBLE…

    All these test have occupied me two weeks.

    Finally in my Company we are working with Excel 2007. Better versions, the older versions.

  8. Jan van Galen

    I was very surpised by the speedgain due to the language setting. Since I have many large sheets which run a considerable amount of time, I decided to take some tests too with my sheets, using your “VBA-Macro_Measuring_Calculation_Time” sheet. In all my cases I did not noticed any speedgain at all. Only your example with the 100K vlookups was way faster. In many of my sheets, calculations are performed by VBA code.
    Thanks for your inspiring work!

    kind regards,
    Jan van Galen
    asus N76VM i7 8Gbyte, Excel 2010/32bit

  9. Ralph Magro

    Very interesting results. Thank you for doing this work and publishing for us to learn.

    Did you try English(USA) against say English (UK) or English (India)?
    Basic numerical formats remain the same but dates differ, so I wondered if they still do some “translation” between them.

  10. Janne

    Thanks for testing!

    My time was 1.5s! Yes 1,5 seconds.

    Either 2016 office is so much faster or then i5-8400 cpu is on another level compared to older cpus.
    Anyone else?

  11. Lyall


    Can you mention the processor difference between the Mac book pro vs the thinkpad. I’ve noticed almost a 1:1 relationship between the processor benchmark scores at cpubenchmark and the calculation performance in Excel. i.e. a I7 6700hq is 70% faster than a I5 6300u

    Also can you compare Excel 2010 vs Excel 2016. My own test show a performance difference of 30% in favour of Excel 2010. Can you confirm?

  12. Dan

    It might be interesting to use similar techniques to compare excel and libre office calc.

  13. Inge

    Thanks, just managed to go from 10 minutes to 30 seconds calculation time. Especially SUMIFS to VLOOKUP made a huge difference. I was using a list of 500k lines to look stuf up 🙂

  14. ExcelFun

    One test you forgot to do is

    100000 lines of 100 columns with random data while say 40% of one column has same data, You filter the data where you have 40% of the data in random rows (non-sequential). Then try

    – Filter the 40% data column then Delete Rows
    – Sort the 40% data column then Filter and finally delete rows

    This will be NIGHT and Day in performance… like HDD vs SSD in performance.

    • Henrik Schiffner

      Thanks for your reply. I’ll keep that in mind and – if I’m eventually do a follow-up study – add this to the test. But please feel free to use the measuring workbook and share your results! 🙂

  15. Pelucchi

    Comparison between 4 comps:

    i3-6100u loaptop 7,8s and after 0365 1809 build time dropped to 0,16s! Now it is only 0,16s.
    i7-4800mq laptop 40-44s
    i7-8650u initially 45-50s (I dont know why it is slower than i7-4800mq). After couple of days use with new computer benchmark time dropped to 2,9s with same office build 1803
    i5-8500 desktop 1,5s with office build 1803

    –>office build 1809 is MUCH faster in this test. Stragely my slowest computer is fastest in this test!

  16. Franck

    Excellent article.
    I also made a few tests switching the Display language from French to English. Gain of roughly 10%.

  17. Stefan

    0,07 seconds

    desktop from 2015 – intel i5 4690k @ 4000 MHz 16 GB of RAM

  18. Earl J

    @Peluchi Interesting results. However, I’m wondering why you chose to use the 1809 build just for the slowest i3 PC, instead of limiting the variation and using the same build across all PCs tested. I’m planning to test this myself when I get home as well.

    Does anyone know if more RAM would affect the performance of these tests?

  19. Matt

    This is very interesting. I was wondering if formulas performed faster if for certain cases you had the first calculation return zero rather then calculate the whole long formula out? For example say you had four columns with really long formulas for the overall rating of a Pitcher, Outfielder, Infielder, or Catcher. Would it perform faster if you surrounded the “Calculate OVR Pitcher” column of formulas with an IF(Postion = “Pitcher”, Calculate OVR Pitcher, 0) instead of just having the Calculate OVR Pitcher formula?

  20. Lorenzo VM

    This may be sacrilegious, but have you any thoughts on a similar test for G***le sheets?

    I’ve but trying to determine how much the client machine’s HW matters vs how much is server side.

    my searches led me here, however the macro does not import.

    If nothing else, a comparison of the two could help.

  21. sugi

    Do you mean : “language English (USA)” ?
    not “Region English (USA)” ?

Leave a comment

%d bloggers like this: