No announcement yet.

DirectX 9 performance is always low.

  • Filter
  • Time
  • Show
Clear All
new posts

  • DirectX 9 performance is always low.

    I recently installed a RTX 2070 Super in my computer and to my surprise, the DirectX 9 test doesn’t produce a high frame rate. DirectX 10, 11 run 350-400 FPS on my 1920x1080 monitor. In DirectX 12, I loose 60% on my score. It would seem the DirectX 9 should be at least 350-400 FPS being it’s much smaller in resolution, but it’s producing a much lower 175 FPS. My 3D score is really being knocked down from it. DirectX 10 and 11 are 99th percentile, DirectX 9 is at 79th percentile. Any idea why DirectX 9 is being slowed by so much?

  • #2
    Maybe your CPU is a bottleneck?


    • #3
      I have considered that as a possibility. My computer has dual Xeon E5-2667 V3 8 core CPU’s, base clock of 3.2 Ghz. Turbo boost varies on how many cores are being utilized. The question becomes if a Passmark CPU score of 23500 is a bottleneck for the DirectX 9 test and its score.
      DirectX 10,11 and 12 certainly aren’t bottlenecked and are producing good results.
      My question is why DirectX 9 is different from these tests


      • #4
        Of all the 3D tests in PerformanceTest, the DirectX9 test is the oldest and runs at the lowest resolution. It is also lightly threaded. So your dual 8 core CPUs count for nothing in gaming. The single threaded rating is more important and the Xeon E5-2667 V3 isn't a great single threaded CPU. Maybe you are running ECC RAM as well, which won't help.
        Dual CPUs also have other problems with NUMA memory access not always being optimal.


        • #5
          Yes the items you have mentioned do effect Passmark's video performance tests, This is unfortunately for myself because I've used Passmark for many years. I also use 3DMark benchmark test and it does produce very high scores with my computer. Firestrike is around 22500 and TimeSpy is around 10900 respectively. This puts my computer at the 95% percentile of all testing results using the very same settings with 3DMark test as Passmark.
          Having tested my computer with the RTX 2070 Super and the Quadro M4000, my Passmark score is almost identical between the two video cards. 3DMark FireStrike test definitely shows a huge difference between the two cards; RTX 2070 Super 10900 vs Quadro M4000 2600.
          Passmark has been around for a long time, I suspect many of the tests are somewhat outdated; especially with the Spectre, Meltdown and MDS patches destroying 2D performance tests. I know how bad these patches are hurting the 2D test scores...


          • #6
            Different software performs differently on different systems. That's not really 'unfortunate', it is just the way it is.
            If we are ranking your machine between the 79th percentile and 99th percentiles, depending on the test, and some other software tells you 95th percentile, I don't really see this as being wrong or inconsistent. I understand you might be a bit annoyed that you got yourself a dual CPU system, thinking that two CPUs must be twice as good as one CPU, then find out that this isn't always the case. It depends enormously on the software if dual CPUs are of any use.

            2D results have no influence on the 3D results. But yes, all those security patches did slow down everyone's computers. Saying that we should release new benchmarks just for the purpose of hiding the performance drop from Windows patches doesn't make sense. It is job of a benchmark to allow measurement of the performance impact.

            Having said that, we are in the process of developing PerformanceTest V10. But our goal wasn't to work around the security patches. Nevertheless we are in the process of overhauling the 2D tests (and memory tests, and CPU tests)