Announcement

Collapse
No announcement yet.

Differences in 2D mark score

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Differences in 2D mark score

    I noticed significant differences in the 2D MARK test, depending on whether I run all tests (CPU, 2D, 3D, MEMORY etc.) or run just the 2D MARK test alone. If I run only 2D MARK, my 2DMARK score is over 1300, if I run all tests, my 2DMARK score is 980. In the first case the test runs noticeably faster (faster scrolling images, pdfs etc.), in the second case everything runs a little slower.The other results (CPU, 3D, MEMORY, etc.) do not differ when I run them separately and when I run them all together. Why the big difference between the 2D MARK results.
    My computer is i5 12600K @5.0GHz, MSI 1050 ti OC, RAM 32GB 3600MHz.


  • #2
    We don't see anything like this on our systems here (2D actually went up a bit).

    2D results depend on the CPU as well so perhaps the system is overheating/throttling a bit after running the CPU tests first.

    Also assuming you are using the latest version (10.2.1001) as support for 12th Gen Intel CPUs was only added in 10.2, if not PerformanceTest will be starting too many CPU threads (20 on this CPU instead of 16) so it could be causing the CPU to be under more load that expected.

    Comment


    • #3
      I'm experiencing the same problem (i7-12700KF, RTX 3080). After much experimentation, I'm convinced it's a bug.

      Launching a fresh session of PerformanceTest 10.2, I ran each 2D test individually multiple times, and it always reported the expected (high) values. Then I clicked the button to run the entire 2D suite (which should be functionally identical to running each test individually). After this point, running any 2D test (individually or as a suite) would report unexpectedly low (by about 50%) values.

      Notice that it's not tied specifically to CPU tests. Running any other test (as a suite) will trigger the bug. And it's only triggered by running a test suite, not individual tests.

      Comment


      • #4
        Might be temperature throttling on your CPU. Running all the tests heats up the CPU (and whole system) more.
        And max CPU clocks speeds can be throttled down a bit once the system gets hot.

        Comment


        • #5
          I don't think it's temperature related. My CPU stays under 40 degrees during the 2D tests. And even if I wait ten minutes after running a suite, the next 2D test will have a bogus low score.

          Comment


          • #6
            These 12th Gen CPUs have power Cores and Efficiency Cores.
            Maybe operating system is (incorrectly) running the tests on the Efficiency Cores some of the time.

            Comment


            • #7
              We've finally been able to reproduce this on an i5-12600K, after the initial run the main PerformanceTest thread was being run on an e-core and not a p-core. Initially we thought it might be a windows thread director issue however we believe we've narrowed it down to an issue in the temperature collection (changing the thread affinity but not correctly setting it back).

              We should have a new build ready for you to test later today.

              Edit: There is a new build ready to test that should now fix this issue.
              Last edited by Tim (PassMark); Mar-29-2022, 06:17 AM.

              Comment


              • #8
                Awesome, thanks for the responsiveness! The test build fixed it!

                Comment


                • #9
                  The fix is now available in the public release version 10.2.1005.

                  Comment


                  • #10
                    sO THERE IS A FIX?that is great news.I was noticing the exact same thing even though my rig is nowhere near 12th gen.In chasing the highest scores (for my models of course) I get hundreds of points above my gpu model recors and Ive also beaten ram highest score (gskill v 3200mhz c 16) However when I run all the tests in order to get them submitted I just cant seem to replicate my individual scores,and I know there is no temp problem.Guess I will download this fix and try again ....thankyou and keep well to all....BPH

                    Comment


                    • #11
                      Fix should only effect CPUs with e-cores and p-cores (i.e. different types of cores with different performance levels). So only intel 12th Gen at the moment (and some ARM CPUs).

                      Comment


                      • #12
                        Thanks Dave,a real pity though,as when running memory benchmark I have my memory (gskill v c16 3200 8gvkb) beating the record (for same model) by quite a margin.However soon as the entire suite of tests are chosen, the score drops +- 50 points, anyway I guess its no big deal but it would be great if you chaps could fix this "bug"...keep well...

                        Comment

                        Working...
                        X