DirectX 9 CPU Benchmark Thread

Discussion in 'General Discussion' started by Thomas Jansen, Jul 22, 2019.

  1. Maarten

    Maarten Member

    Joined:
    Apr 15, 2019
    Ratings:
    +23 / 0 / -0
    It's a cpu instruction which is pretty demanding on your cpu. To my knowledge it's not used a lot. If you do not use AVX you can run a higher clock. To determine if your overclock is stable you run benchmarks to test it, you don't want them to use AVX because that draws a lot of current and causes heat issues.

    If you set a high AVX offset, thereby limiting the frequency of your cpu when avx is used, the clock will collapse to the your cpu frequency minus the offset and causing performance issues. To be sure this does not happen, you can disable AVX in windows.
     
    • Informative Informative x 1
  2. Maarten

    Maarten Member

    Joined:
    Apr 15, 2019
    Ratings:
    +23 / 0 / -0
    Yesterday I've tested a 1080ti in my system and ran the benchmark again. The results between the 1060 and 1080ti were identical. In Raceroom with the same settings, my overhead went up just by 20-25%.

    I'm using a Rift CV1, and with my 1060 I am only able to use super sampling of 1.0. Now with the 1080ti I was able to run super sampling at 2.0 butter smooth.
    So it is true Raceroom does depend on the CPU a lot, but the resolution seems to depend strongly on what GPU you use as well. Also the shadow glitches I had with the 1060 are gone with the 1080ti.
     
  3. Balrog

    Balrog Well-Known Member

    Joined:
    Apr 10, 2015
    Ratings:
    +466 / 0 / -0
    Does anyone have sim-related experience with the latest AGESA (1.0.0.6 / V2) and Ryzen performance? Are the new BIOSes worth bothering?
     
  4. mmmmbeer

    mmmmbeer Member

    Joined:
    Jun 21, 2015
    Ratings:
    +6 / 0 / -0
    So I built a new system in anticipation of the Ryzen 4000's later this year. Have a 3600 in it until then. Tried various setups. I've lost the silicone lottery based on some of the results from the chart on page one. All good jumps from my previous scores though.

    All stock, XMP on , PBO auto yielded average FPS of 332.2
    All stock, XMP on, PBO +200 yielded average FPS of 336.5 (best on page one was 363.9!)
    Best result was a manual OC to 4.4ghz all core which yielded average FPS of 348.8.
     

    Attached Files:

    • Like Like x 1
  5. Larry Foster

    Larry Foster Member

    Joined:
    Apr 14, 2019
    Ratings:
    +13 / 0 / -0
    @mmmmbeer I noticed in your pics that the 2080ti is running 7000 MHz on its memory.
    Is that stock or OC?
     
  6. mmmmbeer

    mmmmbeer Member

    Joined:
    Jun 21, 2015
    Ratings:
    +6 / 0 / -0
    Stock. I can get it up to 8100 before things get flaky.
     
  7. Jason Schofield

    Jason Schofield New Member

    Joined:
    Oct 16, 2019
    Ratings:
    +1 / 0 / -0
    Just upgraded my motherboard so I could enjoy my full 3200 ram speed. x570.
     

    Attached Files:

    • Like Like x 1
  8. Davizmg

    Davizmg New Member

    Joined:
    Sep 29, 2019
    Ratings:
    +2 / 0 / -0
    why my minimum fps are so low??
    • CPU: i9 9900k
    • CPU OC: 5.1 GHz
    • Memory: 32GB(4x8GB) DDR4-3600 MHz
    • Memory timings: CL16-16-16-36
    • GPU: Nvidia Geforce RTX 2080ti
     

    Attached Files:

    • Like Like x 1
    Last edited: Sep 19, 2020
  9. The Iron Wolf

    The Iron Wolf Well-Known Member

    Joined:
    Mar 18, 2015
    Ratings:
    +130 / 0 / -0
    Hello,
    Just built a new system. Still trying to understand why 4000Mhz memory barely runs at 3100Mhz, but I also have another question and thought I'd reach out.
    • CPU: i9 10900KF
    • CPU OC: 5.1 GHz
    • Memory: 32GB(2x16GB) DDR4@3100
    • Memory timings: CL13-13-13-34
    • GPU: Nvidia Geforce RTX 2080ti
    What I don't understand is this:
    If I lock all cores to 5.1Ghz, I get 10k marks. But if I put something like
    3 cores x 5.3,
    3 cores x 5.2
    4 cores x 5.1

    I get 9.3k, fully reproducible, no throttling. This is counterintuitive, anyone has an explanation?
     

    Attached Files:

    • Like Like x 1
    Last edited: Sep 29, 2020
  10. fl0wf1r3

    fl0wf1r3 Well-Known Member

    Joined:
    Apr 24, 2016
    Ratings:
    +235 / 0 / -0
    i9-10900KF it like to take a sledgehammer to crack a nut (raceroom) lol
     
  11. Bull Shark

    Bull Shark Well-Known Member

    Joined:
    Mar 14, 2019
    Ratings:
    +162 / 0 / -0
    well my CPU does not a bad job considering it is “only” a i7 8700k proc.:D

    9A9BCC3F-ECEB-4F66-884E-C73E36EADA3D.png
     
    • Like Like x 1
  12. Thomas Jansen

    Thomas Jansen KW Studios Developer Beta tester

    Joined:
    Apr 5, 2018
    Ratings:
    +562 / 0 / -0
    We've definitely noticed more threads in general do work better in unigine heaven, so unfortunately it's not 100% accurate for performance in RR itself. It just seems like there is a small part of the rendering that can be spread over a ton of threads, SMT on is also better in this for example.
     
  13. fl0wf1r3

    fl0wf1r3 Well-Known Member

    Joined:
    Apr 24, 2016
    Ratings:
    +235 / 0 / -0
    Hello, a new Benchmark of my workstation. AMD Ryzen Threadripper 3990X
     

    Attached Files:

    • Like Like x 2
  14. Balrog

    Balrog Well-Known Member

    Joined:
    Apr 10, 2015
    Ratings:
    +466 / 0 / -0
    Bloody hell, that CPU clearly isn't for single thread tasks, but it's a monster, what are you using it for normally?
     
  15. fl0wf1r3

    fl0wf1r3 Well-Known Member

    Joined:
    Apr 24, 2016
    Ratings:
    +235 / 0 / -0
    right, we are doing images for the automotive industries with mainly vray, a render engine for 3d software like Maya in our case.
     
    • Like Like x 1
  16. Bull Shark

    Bull Shark Well-Known Member

    Joined:
    Mar 14, 2019
    Ratings:
    +162 / 0 / -0
    Cool I worked for more than 17years with Cinema4D and was very happy to see Vray started to support C4D. Good old times. Always get a warm feeling when there is one who speaks about 3D modelling.
    The picture underneath is modeled by me in C4D (in 2011)and rendered with one of the first versions of VrayforC4D.


    3DF1C288-596C-4BDC-B2A2-A09BD46CE324.jpeg

    Sorry, back on topic..


    First I’ll buy a new GFX card probably a 3080 but will wait a bit longer and see what AMD has to offer.
    After that I’m planning a new proc as well although my i7 8700k is still fine for me.
     
    • Like Like x 1
  17. fl0wf1r3

    fl0wf1r3 Well-Known Member

    Joined:
    Apr 24, 2016
    Ratings:
    +235 / 0 / -0
    The AMD Ryzen Threadripper 3990X is only good for rendering. The more cores the better but it does not make Maya any faster its not a software that I would recommend for task like this, doing full cgi images. By the way I just realized that I had the wrong graphics card driver installed, am using the studio driver not the gamer version
     
    • Like Like x 1
  18. Balrog

    Balrog Well-Known Member

    Joined:
    Apr 10, 2015
    Ratings:
    +466 / 0 / -0
    Yeah, the first leaked benchmarks suggest the Ryzen 5000 series will arrive with significantly improved single thread performance, which is pretty tempting, originally I planned to skip this generation, but we'll see.:rolleyes:
     
    • Agree Agree x 1
  19. Bull Shark

    Bull Shark Well-Known Member

    Joined:
    Mar 14, 2019
    Ratings:
    +162 / 0 / -0
    Nice nice. I’m not in a hurry to buy a new CPU or GPU as my system still runs fine. But never say never. That said the most racegames I have do run fine atm.
    First I’ll buy Star Wars :Squadrons. VR and sitting in a XWing is always a +1.
     
  20. The Iron Wolf

    The Iron Wolf Well-Known Member

    Joined:
    Mar 18, 2015
    Ratings:
    +130 / 0 / -0
    the fact that more threads help (a tiny bit) does not surprise me - nVidia driver is heavily multithreaded. You can see that in a debugger. Even GTR2 gains from more threads.

    What I don't understand though is why all cores at 5.1 work faster (in this test) than 3x5.3/3x5.2/3x51. It's like there's some overhead from jumping frequencies.
     
    Last edited: Oct 5, 2020