Announcement

Collapse
No announcement yet.

3D Graphics Test

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

  • 3D Graphics Test

    So something a bit strange is going on. I just built a new system 13900k with RTX 4090. All the 3D benchmarks except for DX12 are running where I think they should be. I'm comparing to my other system which has a 3090Ti. My 3090Ti is averaging about 60fps more than the 4090 in the DX12 test, but is all other tests the 4090 kills the 3090Ti.

    Ideas?

  • #2
    Just to elaborate a little on the two systems. System 1 Gigabyte X570 Master MB, AMD 5950X, 32GB Gskill, EVGA 3090Ti Ultra. System 2 Gigabyte Z790 Master MB, I9 13900K, 64GB GSkill, Gigabyte Aorus Master 4090. Both using MSI 32" monitors.

    In DX12 bench on Passmark system 1 seems to average around 240fps. System 2 averages around 180 fps.

    In DX12 3DMark bench I get the exact opposite (as it should be) where system 1 gets a score of about 5800 vs. system 2 score of 10500

    IS there something going on wrong in the Passmark DX12 benchmark? The 4090 should not be running a full 25-30% slower than my 3090Ti.

    Just wondering if anyone else has also seen these odd results?

    Comment


    • #3
      Distribution of results is below for the 4090 and the 3090ti

      Click image for larger version  Name:	RTX3090ti-DX12.jpg Views:	0 Size:	89.0 KB ID:	53984
      .
      Click image for larger version  Name:	RTX4090-DX12.jpg Views:	0 Size:	87.1 KB ID:	53985
      So for both cards there is a very wide distribution of results in the DX12 test in real world machines, with a very significant overlap.

      The average result however is in line with expectations.
      118fps for the 3090ti
      145fps for the 4090

      The wide distribution means that there are lots of different ways to have both these cards setup badly. Bad cooling, under performing CPU, wrong PCIe slot, different video card drivers, fps locked to monitor refresh rate, low resolution monitor (as this DX12 tests was designed to run at 4K), etc..

      Both your systems are well above the average however.
      For the 3090ti you are literally off the charts above the average (we trim outliers for the charts above to neaten them up).

      So I think the question is not why is your 4090 slow. The question is what did you do to the 3090ti to get it so fast?

      Comment


      • #4
        I'm thinking it's more a driver issue right now as it's not playing "as nice" with Passmark's DX12 test.

        I have two 3090 Cards EVGA 3090 FTW (OC Bios) and EVGA 3090Ti FTW. Both get similar results in Passmark tests. The only test in Passmark 3D testing where my 4090 falls short is the DX12. I really don't think it's a matter of why my 3090 is fast as two cards do the basic same thing. I am also running an older driver on those cards (517.4 which seemed to work well for me.on the 4090 I am running 527.37.

        Again though with 3dMark's DX12 test the 3090's perform at about 1/2 the performance level of the 4090 (Mid 5000 score vs. mid 10000 score). Seems for some reason there is some performance being lost in the Passmark DX12.

        Maybe something to look into?

        Comment


        • #5
          3DMark(Timespy) also includes a CPU test. So some of the higher result you are seeing in 3DMark might be due to the Intel CPU's significantly higher single thread performance.

          But as mentioned you are getting above the expected DX12 result for the 4090. Can't fix something that isn't broken.
          Maybe you can squeeze out a new more percentage points. But there is nothing to indicate anything is broken with the 4090 (except the crazy cooling & power requirements common to all the 4090 cards)



          Comment


          • #6
            I'm using their DX12 benchmark which is called "Speed Way". I believe it is something relatively new, and I think it's more GPU dependent than CPU. I really don't think something is wrong with either of my GPU's. Just saying the 4090, especially with the 13900K should be way outperforming my 3090's in Passmark DX12 test, and it's not.

            Comment


            • #7
              More testing -

              GPU-Z Max GPU Load %

              Passmark DX12 Test 3090Ti 59% highest load

              Passmark DX12 Test 4090 43% highest load

              Passmark DX11 Test 3090Ti 86% highest load

              Passmark DX11 Test 4090 95% highest load

              Passmark DX10 Test 3090Ti 98% highest load

              Passmark DX10 Test 4090 97% highest load

              Passmark DX9 Test 3090Ti 99% highest load

              Passmark DX9 Test 4090 88% highest load

              It looks like the DX12 test is leaving a lot on the table, especially with the 4090 (less than half usage at its highest point)​​​

              Comment


              • #8
                What resolution is your monitor? Are you running G-Sync?

                Comment


                • #9
                  Good call! When I installed the OS on this new system I forgot to change the default refresh rate.

                  Ok, so now things are where I would expect them to be

                  Click image for larger version

Name:	Screenshot_20221208_063053.png
Views:	247
Size:	14.4 KB
ID:	54014

                  Thanks for your help!

                  Comment

                  Working...
                  X