No announcement yet.

Low 3D Mark score

  • Filter
  • Time
  • Show
Clear All
new posts

  • Low 3D Mark score

    Hello, tell me, please, what could be the reason for the low score of the 3D Mark of the RTX 3080 video card?
    At the beginning I had MSI 1060 3 GB graphics card that was included in the test, and I got the following result:
    Then I replaced the graphics card with a GIGABYTE GeForce RTX 3080 GAMING OC 10G (rev. 2.0) (LHR), uninstalled the graphics card driver using Display Driver Uninstaller, and installed the latest GeForce Game Ready 516.59. As a result, the 3D Graphics Mark score was only 15739, although it should be about equal to the average score of 25107 for the RTX 3080. Moreover, the DirectX 9 score turned out to be 2 times lower than the same score for the GTX 1060.

    The detailed score for the RTX 3080 graphics card is as follows.
    Click image for larger version  Name:	Screenshot_2.png Views:	0 Size:	73.6 KB ID:	53222

    The detailed score for the GTX 1060 graphics card is as follows.
    Click image for larger version  Name:	image.png Views:	0 Size:	227.2 KB ID:	53223

    The characteristics of the computer are:
    1) Motherboard: MSI MPG Z690 EDGE WIFI DDR5
    2) CPU: Intel i7-12700
    3) GPU: GIGABYTE GeForce RTX 3080
    4) RAM: Corsair Vengeance [CMK32GX5M2A4800C40] 32GB (DDR5)
    5) Power supply: Seasonic FOCUS GM-650 (SSR-650FM)
    6) SSD: Samsung SSD 980 PRO 250GB

    I also compared the FPS in some games against the FPS that was shown on the same graphics settings in the examples on YouTube. As a result, the FPS coincided.

    I also ran the 3DMark Time Spy test which produced a score of 16902 on the graphics card, which is broadly in line with the average score of 17631. And the processor score was 16280.

    I suppose that the possible reasons for such a low result may be:
    1. The fact that PerformanceTest 10.2 has ended its trial period
    2. PerformanceTest is not fully designed for video cards with LHR.
    3. PerformanceTest considers GPU Hashrate in its evaluation.
    4. The case in the system Windows 10 Pro for Workstations 21H2 (build 19044.1826)
    5. There is not enough power supply to get at least an average of 25107

  • #2

    Q1. No. Benchmark results don't change depending on the status of the trial period (that would be a useless benchmark if they did)

    Q2. LHR refers to throttling the performance of the GPU hardware when running Crypto hashing. Applications don't need to "design" for this. The DirectX9 test isn't doing Crypto hashing. So the LHR crippling should have no effect. But who really knows, there could be a bug in nVidia's code. It was a super dumb idea by nVidia as it was worked-around by the Crypto miners and it was just one more dumb feature to go wrong.

    Q3. The GPU compute test comes close to this. But as far as we are aware LHR doesn't effect any test. And your results in compute look OK.

    Q4. Windows 10 for Workstation is the same as normal Win10, except you can use more RAM (6TB) and four socket motherboards. But the i7 CPUs don't support this anyway.

    Q5. Very unlikely.

    More likely possibilities are,
    • Bad cooling / dust leading to thermal throttling
    • Thermal paste / pads have gone dry and heat sink isn't working as well as it might
    • GSync limiting frame rates to the monitors refresh rate
    • If this was an ex-crypto mining card maybe it was underclocked?


    • #3
      Thanks for the help. As regards these points:
      The cooling in the computer is quite good, and the dust has not appeared yet.
      The thermal paste/pads are supposed to be ok, because as soon as I bought this new video card, I immediately ran tests. Also, I didn't overclock the card.
      My Acer R271 bmid monitor does not support GSync.
      I also ran FurMark Test and the card works stably in this test. In this case, the temperatures become constant: the GPU temperature is 70 degrees Celsius, and the memory temperature is 84 degrees Celsius.
      In addition, it is not clear why the score is low on the directx 9 test.
      Last edited by Kazaams; Jul-28-2022, 03:19 PM.


      • #4
        Next most likely possibility is 3rd party software.
        Game overlays, screen recording software, remote desktop solutions, frame rate monitors, etc....

        Then check that DirectX9 is actually running on 3080. When you have two graphics cards in the system, the software to pick the best card can get confused. Maybe the test is running on the iGPU in the CPU and not on the 30380?

        Then check device driver settings. There are a bunch of settings that can cause performance issues.


        • #5
          1) I made sure that I have everything listed (Game overlays, screen recording software, remote desktop solutions, frame rate monitors) either disabled or missing.
          2) I also disabled the intel graphics card in device manager just in case.
          3) Additionally, in the Nvidia control panel, I configured PerformanceTest64.exe to use only the RTX 3080
          4) Once again I installed DirectX.
          Applying all these points, the result on the directx 9 test was just as low.
          Moreover, I noticed that csrss.exe uses about 1.5% of the GPU when moving the mouse. Although before the GPU tests, for some reason it applied 10-15% of the GPU also when moving the mouse.


          • #6
            Maybe a clean Windows install?
            Or try the same video card in another machine?