DirectX 9 CPU Benchmark Thread

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

  1. Norbi930523

    Norbi930523 New Member

    Joined:
    Oct 30, 2021
    Ratings:
    +3 / 0 / -0
    I checked the temperatures, same results for both bad and good scores:
    - CPU: consistently 44-46°C during the whole benchmark
    - GPU: started at 37°C, slowly increased to 40 by the end of the benchmark
     
  2. Nir Tal

    Nir Tal Well-Known Member

    Joined:
    Aug 29, 2016
    Ratings:
    +53 / 0 / -0
    I suggest opening new thread about your performance issue as more people would be able to help.
    Look to me as the cpu is under-performing for some reason. Would suggest to benchmark only the cpu with Cinebench as its not using the GPU and may isolate the problem.
     
  3. Ablaze

    Ablaze Well-Known Member

    Joined:
    Mar 16, 2018
    Ratings:
    +120 / 0 / -0
    @Norbi930523 If you use GPU-Z to show your PCI-E bus interface, what is displayed there?

    [​IMG]
     
    • Like Like x 2
  4. Norbi930523

    Norbi930523 New Member

    Joined:
    Oct 30, 2021
    Ratings:
    +3 / 0 / -0
    @Ablaze it says: PCIe x16 3.0 @ x2 1.1, which sounds bad :D (I also saw @ x2 3.0). I tried setting PCIe to Gen3 specifically, but didn't change anything. Other solutions I found is cleaning out the slots or might be some bent pins. The card itself is in the PCIe x16 slot.
     
  5. Ablaze

    Ablaze Well-Known Member

    Joined:
    Mar 16, 2018
    Ratings:
    +120 / 0 / -0
    @Norbi930523 Click on the small ? to the right of the bus interface field. This will produce a load on your GPU and will show the correct bus speed. If it's not showing x16 3.0 you have the reason for your perforamce problems. If that's the case make absolutely sure that your card is in the very first PCIe 16 slot of your mainboard, just directly below the CPU socket. The PCIe speed for this slot must be set to auto or Gen 3 in your BIOS.

    A friend of mine had the exact same issues with his PC a few days ago and it turned out he inserted the GPU in the wrong PCI slot so the card was only running with x4 1.1. The performace increase after the GPU was running with x16 3.0 was to the moon and beyond. :D
     
  6. Norbi930523

    Norbi930523 New Member

    Joined:
    Oct 30, 2021
    Ratings:
    +3 / 0 / -0
    x2 is the most that I get. But from what I've read so far, the link width (x1, x2, x4, etc) doesn't depend on the load on the GPU, but that's a different topic :D I tried reseating the GPU, resetting CMOS, cleaning the PCIe slot (I have only one), cleaning the GPU, still x2. Fortunately, I still have my old GPU so I put it back and GPU-Z shows x16 link width for it. One catch: I had to connect the monitor to the integrated graphics HDMI slot, as my old GPU doesn't have HDMI, only DVI and mini-HDMI, and I'm not sure how this affects the results - although I checked my new GPU the same way (i.e. connected to integrated graphics), and GPU-Z still showed the correct data (and x2). So since my old GPU has x16 link width, that makes me believe that the issue is with the GPU itself.
     
  7. Norbi930523

    Norbi930523 New Member

    Joined:
    Oct 30, 2021
    Ratings:
    +3 / 0 / -0
    Finally! I've got a replacement card (still GTX 1660 Ti), and now this one operates at x16.

    While running Unigine:
    - Average FPS: 233 vs previous best 164 (worst was around 81)
    - GPU load: around 40% vs previous 98%
    - Bus interface load: 15-20% vs previous 50% (when I got good FPS) and 85% (when I got bad FPS)

    Thank you all for all the tips and advice! :)
     

    Attached Files:

    • Like Like x 3
    • Winner Winner x 1
  8. mclaren777

    mclaren777 Member

    Joined:
    Feb 14, 2021
    Ratings:
    +7 / 0 / -0
    I've spent ~20 hours trying to troubleshoot my issue and I feel like I'm going crazy – this is all I can think about lately.

    I bought an RTX 3080 over the summer to replace my GTX 1080, but didn't notice any improvement in RaceRoom's framerate. Everything appeared to be working normally in other games (MS Flight Sim runs great) and I often heard the claim that R3E is fundamentally hampered because it doesn't use CPU cores very well. I was content to move on and hope for a future DX11 update, but I recently bought a gaming computer for my son and it opened up a can of worms. His ancient GTX 970 is able to run R3E at full GPU load (screenshot), whereas my GPU never wants to go above 50% (I can get closer to 60% by using DXVK – YouTube video).

    Anyway, I discovered this thread and I decided to run the Unigine benchmark on both machines today, and my son's computer beat mine yet again. I don't understand why my computer seemingly struggles with DX9 rendering, but I'm hoping someone here might know what's going on.

    • CPU: Intel 8700 (stock)
    • Memory: 32GB DDR4-2666 MHz Dual
    • Memory timings: CL15-16-16-35
    • GPU: Nvidia GeForce RTX 3080
    [​IMG]

    • CPU: Intel 9700F (stock)
    • Memory: 16GB DDR4-2666 MHz Dual
    • Memory timings: CL19-19-19-43
    • GPU: Nvidia GeForce GTX 970
    [​IMG]
     
  9. Vale

    Vale Well-Known Member

    Joined:
    Jul 4, 2019
    Ratings:
    +278 / 0 / -0
    If they are both desktops then why not swap around the graphics cards and repeat the tests to see what happens?
     
  10. Maskerader

    Maskerader Well-Known Member

    Joined:
    Oct 6, 2019
    Ratings:
    +352 / 0 / -0
    @mclaren777

    A few things I noticed.
    1) 9700F does have a slightly higher single-thread performance than 8700, so a bit higher score is expected because this test with Unigine is configured to focus on that. GPU is barely loaded, so shouldn't be much difference between 3080 and 970. (But maybe you could find something if you swap GPUs, as Vale suggested.)
    8700: https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7-8700+@+3.20GHz&id=3099
    9700F: https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7-9700F+@+3.00GHz&id=3465

    However, in the tests the difference in average FPS is not "slightly higher", but almost 20%. Could be there are other factors too.

    2)
    His CPU is also 100% loaded on all 8 cores which is absolutely not what should happen with this game normally. I suspect there's something else going on, and if it is, it might also affect GPU load. Is this with DX wrapper too?

    For comparison, on your PC the CPU sees much lower load, basically only two cores are doing the job:
    3) I wonder what FPS both systems get, on the same resolution.
     
    Last edited: Dec 1, 2021
  11. Nir Tal

    Nir Tal Well-Known Member

    Joined:
    Aug 29, 2016
    Ratings:
    +53 / 0 / -0
    Following @Maskerader answer, are both PCs on same monitor resolution and game graphic settings ?
     
  12. mclaren777

    mclaren777 Member

    Joined:
    Feb 14, 2021
    Ratings:
    +7 / 0 / -0
  13. Vale

    Vale Well-Known Member

    Joined:
    Jul 4, 2019
    Ratings:
    +278 / 0 / -0
    I think this may be the issue too. The benchmarks show different screen resolutions higher on the slower fps system) and in game settings may also be different so ideally copy and paste the settings from my docs from one system to another and repeat the tests using the same monitor for both pcs at the same default res.

    On another note, Nvidia drivers are now on v 397 so could be worth updating, especially on the newer card system.
     
    Last edited: Dec 2, 2021
  14. Maskerader

    Maskerader Well-Known Member

    Joined:
    Oct 6, 2019
    Ratings:
    +352 / 0 / -0
    Still the same, kinda?

    R3E is a program that is mostly limited by your single thread performance (until you hit very high screen resolutions and your GPU becomes a bottleneck instead). Your son's PC CPU is more powerful than yours. This is also reflected in the benchmarks you linked.

    Then, as Vale said, if these PCs run different resolution, that would put more load on your PC than on your son's one.

    Basically, I would expect your son's PC to get better FPS in these circumstances.

    Also, there could be one thing that I noticed when I did my tests a year ago: even if running the same replay, my FPS were different each time I load it. That difference can go beyond 10% between different runs, and it's not like "earlier runs are better, later runs are worse" or something, it looks like random. So one run is not enough to judge how much higher or lower your FPS are, exactly. I don't know the reason for that, though.
     
  15. mclaren777

    mclaren777 Member

    Joined:
    Feb 14, 2021
    Ratings:
    +7 / 0 / -0
    Regarding UserBenchmark, the raw stats are very similar between both computers with the exception of GPU performance, which massively favors my machine.

    But I find the video comparison to be really damning. That's the same replay file, using identical video settings, run on the same 1080p240 monitor, yet there are times where my son's computer is 30+ fps ahead of mine. R3E refuses to use my system's resources and I'm determined to figure out why.
     
  16. Maskerader

    Maskerader Well-Known Member

    Joined:
    Oct 6, 2019
    Ratings:
    +352 / 0 / -0
    The fact that R3E engine is bad in multithreading and is limited by single-thread performance, is that not an answer for you? If single-thread performance is the bottleneck, it's doesn't matter how good your other components are, they won't be loaded to 100%.

    P.S. Memory scores are also a bit higher on your son's PC. It really looks like your PC is weaker in components that are important for R3E and is stronger in components (GPU) that aren't bottlenecks and don't affect the game performance much.
     
  17. mclaren777

    mclaren777 Member

    Joined:
    Feb 14, 2021
    Ratings:
    +7 / 0 / -0
    After spending 20+ hours troubleshooting my poor performance in RaceRoom, I finally found a solution. It wasn't overclocking my RAM or updating my GPU driver – it was simply starting a new profile!

    If you have a strong computer but you're struggling to see robust performance from this game, here's my advice: open RaceRoom and make a video where you scroll through all of the settings in the game (input bindings, graphics settings, etc). Exit out, then go to "My Documents: My Games: SimBin" and change the name of that entire folder (eg, SimBin Old). This will force the game to create a new profile at launch. Then manually replicate the settings you had before by referencing your video and hopefully you'll see meaningful gains like I did.

     
    • Like Like x 1
  18. Vale

    Vale Well-Known Member

    Joined:
    Jul 4, 2019
    Ratings:
    +278 / 0 / -0
    Good news. Did you go through the two sets of settings to see what changed? I would say there was either some graphics detail shader setting or a controller profile slowing things down. Are all settings now exactly the same as on your son´s pc?
     
  19. mclaren777

    mclaren777 Member

    Joined:
    Feb 14, 2021
    Ratings:
    +7 / 0 / -0
    I haven't compared the internal files yet, but that might be a fun exercise for this weekend.

    Regarding my son's computer, we were always running the same graphics settings and resolution for these comparison videos. Here's the final showdown...

     
    • Like Like x 2
  20. Yuval Rosen

    Yuval Rosen Well-Known Member

    Joined:
    Jun 18, 2021
    Ratings:
    +49 / 0 / -0
    Here are my results:
    • CPU: Intel Core i7-4790
    • CPU OC: Stock
    • Memory: 16GB DDR3-1600 MHz
    • Memory timings: CL11-11-11-28-208-1T
    • GPU: Nvidia GeForce RTX 3070 (Founders Edition)
    upload_2022-1-4_21-32-51.png
    upload_2022-1-4_21-32-38.png
    upload_2022-1-4_21-32-19.png