MP5,1 CPU Upgrade to x5690, GeekBench results low

Discussion in 'Mac Pro' started by diavolo770, Jan 7, 2019.

  1. diavolo770 macrumors newbie

    diavolo770

    Joined:
    Sep 25, 2007
    #1
    I just upgraded my 2012 Mac Pro to dual x5690's from the stock E5645's. I've been watching YouTube videos documenting the results and everything I've seen shows results around 25,000 (give or take) on the multi-core score.

    Here's my 'before' Geekbench score:

    https://browser.geekbench.com/v4/cpu/11006212

    and here's the 'after':

    https://browser.geekbench.com/v4/cpu/11557092

    I'm only seeing an improvement of between 500-600 points on both the single and multi core scores.

    I didn't get a Cinebench score before I upgraded but the post-upgrade score comes out to 1513 - 1612 across half a dozen tests.

    I'm running the latest version of GeekBench with Mojave 10.14.2 (both pre and post upgrade) so my last-luster performance improvement is both disappointing and not making a whole lot of sense. I realize synthetic benchmarks aren't the best representation of real-world performance so I'm doing some other CPU-intensive tasks to see if I can "feel" a marked improvement.

    Can anyone offer some insight to why my performance hardly improved?
     
  2. t8er8 macrumors regular

    t8er8

    Joined:
    Dec 4, 2017
    Location:
    Quebec, Canada
    #2
    Are you sure there wasnt any background tasks running like backups or anything else open while you were doing these new tests, I always find I get the highest score after a fresh reboot on a geekbench test, although this minor improvement is kind of an anomaly ive never seen.
     
  3. h9826790 macrumors G5

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #3
    The score looks very normal to me
     
  4. t8er8 macrumors regular

    t8er8

    Joined:
    Dec 4, 2017
    Location:
    Quebec, Canada
    #4
    You were on westmere 6 core already, you were basically only gaining a higher clock so it might not be that bad
     
  5. diavolo770 thread starter macrumors newbie

    diavolo770

    Joined:
    Sep 25, 2007
    #5
    I shut down as many background tasks as possible and disabled as much as I could such as Creative Cloud sync, OneDrive and Google Drive background sync, Time Machine, etc... before I ran the bench the first time. I'm using Monity to watch the CPU temps and fan speeds which all seemed to be at reasonable level at idle and while under load.

    So, after about an hour sitting at idle while I was researching the problem, I ran Geekbench again and the first test after sitting got me a multi-score of 26,354 and one more yielded 25,314. I've rebooted a couple times and re-run Geekbench with similar results.

    Initially I thought that maybe I didn't give it enough thermal paste and it was throttling but the temps never indicated that. Letting it "settle in" doesn't seem like a very scientific explanation so I'm going to keep digging into it.
     
  6. Fooze macrumors member

    Joined:
    Mar 1, 2016
    #6
    It looks like you are using 64gb of RAM. Are they configured as 8x8GB stocks? Try removing the sticks in slot 4 and 8. Your Geekbench score will go up significantly, but real world performance will be less noticeable. Cinebench score is not affected by the number of RAM sticks used.
     
  7. aslowdodge macrumors member

    Joined:
    Nov 12, 2012
    #7
    Out of curiosity, what does removing ram from slots 4 and 8 do?
     
  8. Fooze macrumors member

    Joined:
    Mar 1, 2016
    #8
    These Xeon Westmere X5600 series processors have a triple channel memory controller. Slots 3/4 and 7/8 are shared. This drastically affects the scores in Geekbench which has several tests that max out the memory bandwidth. Real world performance isn't nearly as drastic.

    Here's one of many posts where this topic is discussed.
    https://forums.macrumors.com/threads/triple-channel-memory.1581022/
     
  9. pl1984 macrumors 68020

    Joined:
    Oct 31, 2017
    #9
    It does seem odd that a 40% clock increase would result in a 4% increase in multithreaded core score. I did notice you used two different versions of GB between the scores. Perhaps that might help explain the low increase in score?
     
  10. h9826790 macrumors G5

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #10
    GB4 multithread score is non linear, super hard to understand and compare.
     
  11. pl1984 macrumors 68020

    Joined:
    Oct 31, 2017
    #11
    I feel a 40% increase in clock speed should result in more than a 4% increase in multithread GB score. Even the single thread score is only 20% higher.
     
  12. hugoZA macrumors member

    Joined:
    Jun 4, 2014
    #12
  13. h9826790 macrumors G5

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #13
    OP’s RAM is not in the optimum triple channel config. That has a huge impact in GB4 score.
    --- Post Merged, Jan 8, 2019 ---
    Anyway, IMO, GB4 is not that “stable” as GB3.

    In GB3, we can really predict the score and the result usually fall within that range. Even though RAM config or background can still affect the score, but just few %. Won’t make the user “panic”.

    But in GB4, just little bit deviate from the optimum situation can result in a very large penalty in score. Definitely not in a linear scale. Really hard to understand and compare the result.
     
  14. pl1984 macrumors 68020

    Joined:
    Oct 31, 2017
    #14
    It may not be the optimal configuration however I would expect, all else being equal, a 40% increase in clock speed would result in more than a mere 4% increase in multithreaded score. Wouldn't it be ironic if he configured the memory to its optimal state and he observed more than a 4% increase in multithreaded score?
    I'm no fan of GB but one thing it should be reasonable for is to measure performance changes due to changes made to the same system. If a 40% increase in processor clock isn't represented by more than a 4% increase in multithreaded score than GB is a completely useless benchmark.
     
  15. h9826790 macrumors G5

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #15
    IMO, it is quite useless now indeed.
     
  16. Paradiseapple macrumors regular

    Joined:
    Aug 18, 2010
    Location:
    Germany
    #16
    T
    hank you for showing me this. My trader says the same: my score is normal
     

Share This Page

15 January 7, 2019