No announcement yet.

3 problems with PerformanceTest

  • Filter
  • Time
  • Show
Clear All
new posts

  • 3 problems with PerformanceTest

    I used PerformanceTest a lot a year ago when tuning/optimizing my new 5950X platform. I had PerformanceTest 10.1 1003 installed, and I still have it installed today. In this version, I currently get up to 53664 (5950X without OC, only undervolting curve opt. and slightly changed TDC) and in the new 10.2 1012 about 53167. Something has been changed in the algorithms, you can see. As a result, new scores are compared online with old versions where the same PC scored higher!

    10.1 build 1003 test Medium
    Click image for larger version

Name:	image.png
Views:	157
Size:	418.7 KB
ID:	53874

    10.2 1012 - Medium
    Click image for larger version

Name:	image.png
Views:	118
Size:	162.2 KB
ID:	53875

    The second problem is after getting the total score of the entire PC system, and clicking View online comparsion charts, it takes me to the main page, both the old version and the new one. Something's broken and you can't see how your score compares to others, and you can't see your place in the world!

    Third problem.
    There is also no comparison to other PC systems with the same processor. Currently, when I enter the New Desktop ranking, my processor score points to the 12th place in the world in the Overclocked section, 19th place.. but this is also nonsense because the overclocked 5950X scores over 54,000 but they are missing in these rankings.
    When I click By socket - and select 5950X, these results above 54,000 are missing and the graph shows that my processor is out of scale!

    4rt weird thing is SSD score.
    The fact that it differs in version 10.1 and 10.2 is one thing, but on your website, a lot of people on this 980 PRO 1TB disk achieved scores of over 56,000 and even over 58,000! Even on a clean disk with the latest drivers, a super clean system, I was getting max 43,000!​

    Attached Files

  • #2
    Issue #1
    You can find a complete list of changes between these releases here

    There was no changes to the x86 CPU test. There was a small change to the ARM CPU test, but you aren't using an ARM CPU.

    But you talking about a 0.9% change in the overall score. Which is likely below the margin for error. You really need to to 10 runs of each to get something statistically valid.

    Issue #2 is the site with comparisons of different CPUs.

    But if you upload your result, then you should also get a link just for your particular result. It will be in this format,

    You can also compare your result with others directly in the software without uploading.

    Issue #3
    We have over 6500 benchmark results from the 5950X CPU that you can compare with. See my screen shot below. Your score is about 35th for this CPU. So I am not exactly sure what web page or screen in the software you are referring to here. For some graphs / screens we filter outliers. Can you post a link or screen shot.

    Issue #4
    I don't think you can attribute a good CPU benchmark to having a 980 1TB SSD. Doesn't really make sense. More likely it is just co-incidence. This was a popular high end SSD when the 5950X was popular. As you can see in the screen shot below, there is a wide variety of mass storage devices used with the top 100 5950X CPU results. No doubt Samsung makes popular drives however.

    Screen shot showing top 100 5950X baselines (no filter for duplicates or overclocking, etc..)
    Click image for larger version

Name:	5950Xbaselines.png
Views:	137
Size:	475.5 KB
ID:	53880