Announcement

Collapse
No announcement yet.

Low 2D Performance scores

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

  • #31
    A few days ago I was benchmarking my new system and was getting 630 for 2D.

    Now I can only get about 450. I cant see what has changed on my system. All other scores are as good as before.
    Drivers haven't changed. No updates.
    It's a 3090 and is shown way below where it should be.
    Model average is 990? What could be set wrong that has lowered 2D so drastically?

    Thanks much

    Comment


    • #32
      Originally posted by maddabbo View Post
      A few days ago I was benchmarking my new system and was getting 630 for 2D.

      Now I can only get about 450. I cant see what has changed on my system. All other scores are as good as before.
      Drivers haven't changed. No updates.
      It's a 3090 and is shown way below where it should be.
      Model average is 990? What could be set wrong that has lowered 2D so drastically?

      Thanks much
      It could be something running in the brackground that may have reduced the score.

      As for the overall low score, could be that your particular GPU driver version can result in lower scores. Or when used in combination with something else. (particular hardware, or overclocking settings, or some 3rd party software).

      Comment


      • #33
        Today i tested again my Commputer fairly recent Ryzen System on Windows 7 64Bit with Passmark 9 and Passmark 10;
        as one can see, on Passmark 9 the 2D Test gives 5 Stars, and on Passmark 10 "0" Stars, with the same graphics card; it would be nice if the software could factor in some missing API, as it seems the low 2D number weighs down totally the whole Score;
        Thanks.

        PSMark09

        Click image for larger version

Name:	passmark10T400NOSPECTRE.png
Views:	88
Size:	73.3 KB
ID:	52631

        Attached Files

        Comment


        • #34
          Windows 7 market share is now under 1% of our users and falling each month.

          Graphs are here,
          https://www.pcbenchmarks.net/os-marketshare.html

          We aren't going to go back and try and improve things for Win7 at this point for PerformanceTest V10. Win7 is missing some of the more modern graphical features, like in built PDF handling and DirectX features that are in Win10.

          Maybe just compare your machine using other PerformanceTest V9 results for an apples to apples comparision.

          Comment


          • #35
            There just has to be something weird going on with the 2D Performance Scores, what else is needed to figure this out? Game Mode On, 3D GPU Acceleration on in Windows 11.


            CPU Type 14-Core Intel Core i9-7940X, 4500 MHz (45 x 100) Motherboard Name Asus Prime X299-A II (2 PCI-E x1, 1 PCI-E x4, 3 PCI-E x16, 3 M.2, 8 DDR4 DIMM, Audio, Gigabit LAN) Motherboard Chipset Intel Kaby Point X299, Intel Skylake-X System Memory 65217 MB (DDR4 SDRAM)
            nVIDIA GeForce RTX 3080 Ti Monitor Generic PnP Monitor [NoDB] (29121) Monitor HDMI88 (28121605)
            Click image for larger version

Name:	2D_Graphics.jpg
Views:	73
Size:	105.8 KB
ID:	52659

            Comment


            • #36
              updated with screenshots.

              Comment


              • #37
                EverStaR, the results aren't all that bad for a 5 year old CPU. Not great I agree. But not truly awful either.

                The i9-7940X was a super hot (temperature) CPU, even before any overclocking. Makes me wonder if the overclocking is hurting rather than helping.

                Comment


                • #38
                  Originally posted by David (PassMark) View Post
                  Windows 7 market share is now under 1% of our users and falling each month.

                  Graphs are here,
                  https://www.pcbenchmarks.net/os-marketshare.html

                  We aren't going to go back and try and improve things for Win7 at this point for PerformanceTest V10. Win7 is missing some of the more modern graphical features, like in built PDF handling and DirectX features that are in Win10.

                  Maybe just compare your machine using other PerformanceTest V9 results for an apples to apples comparision.
                  I have a similar issue. I have a dual boot Windows 7/10 Zotac Zbox with an intel Core i7-4770T. In most respects, and . This has an integrated intel HD 4600. Although over 8 years old, it is quite adequate for most of my computing. Under Windows 7 the 2D mark is in the range 11 - 17 , whereas with Windows 10 it is 391 - 394.
                  Click image for larger version

Name:	benchmark_win7.png
Views:	21
Size:	110.6 KB
ID:	53027
                  win10
                  Windows 10 result


                  The model average is 313. Comparing individual tests, as you can see, the Win7 results for Windows interfaces is twice as high,but image rendering is 50 x slower and image filters and direct 2D are less than half the speed.

                  I have wound back the drivers, (and later re-installed the latest one), checked that spectre and meltdown are disabled, run dxdiag, etc nothing makes a significant difference to the video performance under Windows 7.

                  a) but how much of the difference I'm seeing here is likely to be due to the Tests themselves as they are no longer really designed for Windows 7 ?
                  b) has anyone got any ideas what might be causing the windows 7 slow-down - I believe that it is (at least in part) real?




                  Comment


                  • #39
                    The Tests [in PerformanceTest V10] themselves as they are no longer really designed for Windows 7 ?
                    @jerrytaff,

                    Yes. As noted above, PT V10 uses Directx features and PDF rendering features that are only in Win10 and 11.

                    But this doesn't totally explain the bad Win7 result you posted above. For example the Image Filters test uses the GDI+ Windows programming interface and DirectX11 interface.
                    GDI+ was added to Windows in Windows XP. While DirectX 11.0 was added to Windows in the Vista release.

                    But to more fully use DirectX11 (version 11.1) in Win7 you need the Win7 Platform update which pulled some of the Win8 functionality back into Win7. DX11.1 added a whole bunch of new hardware accelerated features. But only partial 11.1 support was back-ported to Win7.

                    And of course there was Direct11.2, 11.3, 11.4 and DirectX12 features as well. All missing from Win7.

                    From a benchmarking point of view it is a difficult thing. Do we only use features from 20 years ago in Windows XP? Or do we use the modern programming interfaces as they become available in Windows, but making the older machines look worse than they are.

                    Again, if you want to benchmark XP/ Vista / Win7 graphical performance and compare results, then maybe stick with PerformanceTest V8 or V9 results.
                    (note that these issues don't effect RAM, CPU or Disk performance where the programming interfaces haven't really changed in 20 years).

                    Comment


                    • #40
                      David (PassMark)

                      I appreciate that you can't compare older compare systems with those with new capabilities, and I'm not suggesting that you shouldn't be expanding the tests to cover them. However, I would have hoped that they could be kept as separate sets of test tests - like the pdf rendering, and that a machine should be able to conduct a complete set or not at all.

                      I guess that what I'm asking is - taking the image rendering tests as an example, is the score of 0.7 a penalty because it couldn't run the whole set, or did it run the whole set, very slowly? Before reading your comments, I'd assumed it was the latter, but now I just don't know.

                      It's important to me because I'm trying to establish what exactly is wrong with my Windows 7 system - I do believe that there is an issue as it has struggled to process video and sound when on Zoom - I had originally thought that was likely to be to do with the webcam drivers. However, the webcam was not attached when I was performing the tests.

                      I've followed the links and tried installing the Platform update, and received the message "Update for Windows (KN267083 is already installed on this computer).As it was already installed it doesn't explain the poor results. I believe the system incorporates every windows 7 update available. (Even though beyond end of life for MS support, I'm still getting regular security updates)

                      As an aside, do you have any comments as to why the Windows Interface is so much faster on Win 7 compared with Win 10? It even beats my new Rysen 5 5600G system running Win11 on that particualr set of tests.
                      Last edited by jerrytaff; 06-20-2022, 02:03 AM.

                      Comment


                      • #41
                        taking the image rendering tests as an example, is the score of 0.7 a penalty because it couldn't run the whole set, or did it run the whole set, very slowly?
                        It isn't clear from your post. It could be either. It should be visible on screen however. The different steps of the test (e.g. rotation, grey scale filter, etc..) are run twice. Once with the GDI+ interface and once with the DX11 interface with hardware acceleration.

                        Typically DX11 is like >10 times faster than GDI+ for the same operations.

                        So it should be obvious if they are run or not and the DX11 test steps should be visible faster (a blur of speed compared to GDI)

                        If you watch the window title you can see this.
                        e.g
                        Click image for larger version  Name:	DX11-Interface.png Views:	0 Size:	200.8 KB ID:	53038

                        why the Windows Interface is so much faster on Win 7
                        Things like transparency, rounded window corners, gradient shading and just the quality of the code the programmer has written all effect the UI speed.
                        I haven't had a detailed look, so I don't know the exact answer.

                        Comment


                        • #42
                          Thanks for the reply. Watching while running the benchmark, it is obvious that it is only running a small part of the test under Win7 compared to Win10. I guess it is probably safe to assume then that the score is heavily biassed to the very much faster DX11 tests that can't be run under Win7.

                          So, getting back to my comments about compatibility with older systems, it is a great pity that it isn't possible to drill down into finer granularity. e.g. to see the results sub-test by sub-test, displaying the GDI+ and the DX11 test results separately. If that was possible, then those running older systems would be able to conduct sensible comparisons when upgrading. It would also resolve your dilemma as to how to maintain some kind of backward compatibility with older systems without constraining the development of support for new features, to everyone's satisfaction. I imagine it would be relatively easy to implement such a change to the software, especially if the extra info is only available to the local user, and not part of ther upload to your database of devices and results.

                          It would also help users and developers in understanding the impact of the OS, which would be quite helpful generally - e.g. if a system gives a different benchmark after a Microsoft update, a driver update, or perhaps when switching between Win10 and Win 11, it would be far easier to understand which sub-system had been affected, making it easier to identify or develop fixes.

                          Comment


                          • #43
                            that the score is heavily biassed to the very much faster DX11 tests
                            Really depends on your hardware and video card drivers as to how good the hardware acceleration is.

                            it isn't possible to drill down into finer granularity. e.g. to see the results sub-test by sub-test, displaying the GDI+ and the DX11 test results separately
                            Pretty sure we record them in the debug log. But no one will see them in that log.

                            There is a message that pops up if DX11 is missing entirely, so it seems the half implemented DX11 is the issue.

                            Windows 7 market share is now under 0.8% of our users and falling each month.

                            Graphs are here,
                            https://www.pcbenchmarks.net/os-marketshare.html

                            So it doesn't make sense to implement something that 99% of people don't need.

                            Plus it is also possible to make the argument that the numbers are in fact a good reflection of the actual performance of the Win7 machine. Modern software built to use DirectX12 & 11 (and fall back to using GDI otherwise) will potentially run pretty bad on Win7, or not run at all.

                            Comment

                            Working...
                            X