Fix game stutter on Win 10 1703-1809

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Mott, Apr 3, 2018.

  1. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    So you are getting better I/O now?
     
  2. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    2,517
    Likes Received:
    241
    GPU:
    MSI 970 Gaming 4G
    Yes, try yourself, run all stock settings and BM using CrystalDiskMark, then run InSpectre and disable them and rerun BM.
    We should see better random access numbers.
    Note there was one BM that went down a bit if I remember with everything disabled in InSpectre.
     
  3. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    Could be due to the newer microcode making things a bit slower, I am actually using an older one that doesn't have any Spectre&Meltdown patch forced. Thanks for the info, its good to know that you do get good I/O again by disabling the mitigations. Or also could be margin of error in the benchmark.
     
    Last edited: Apr 10, 2020
  4. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    2,517
    Likes Received:
    241
    GPU:
    MSI 970 Gaming 4G
    Ok, I haven't done a comparison in a long time.
    Here 1909 results with the same micro-code OS wise, my bios is before any of this came out. i5-3570k @ 4.7ghz and 850evo SSD.

    CrystalDiskMark
    100mb test

    Meltdown and Spectre (with Retpoline) enabled

    * MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
    * KB = 1000 bytes, KiB = 1024 bytes

    Sequential Read (Q= 32,T= 1) : 549.894 MB/s
    Sequential Write (Q= 32,T= 1) : 520.240 MB/s
    Random Read 4KiB (Q= 32,T= 1) : 284.690 MB/s [ 69504.4 IOPS]
    Random Write 4KiB (Q= 32,T= 1) : 253.647 MB/s [ 61925.5 IOPS]
    Sequential Read (T= 1) : 505.202 MB/s
    Sequential Write (T= 1) : 490.169 MB/s
    Random Read 4KiB (Q= 1,T= 1) : 29.997 MB/s [ 7323.5 IOPS]
    Random Write 4KiB (Q= 1,T= 1) : 96.351 MB/s [ 23523.2 IOPS]

    Meltdown and Spectre (with Retpoline) disabled

    * MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
    * KB = 1000 bytes, KiB = 1024 bytes

    Sequential Read (Q= 32,T= 1) : 550.810 MB/s
    Sequential Write (Q= 32,T= 1) : 527.540 MB/s
    Random Read 4KiB (Q= 32,T= 1) : 300.890 MB/s [ 73459.5 IOPS]
    Random Write 4KiB (Q= 32,T= 1) : 267.700 MB/s [ 65356.4 IOPS]
    Sequential Read (T= 1) : 506.541 MB/s
    Sequential Write (T= 1) : 492.619 MB/s
    Random Read 4KiB (Q= 1,T= 1) : 33.656 MB/s [ 8216.8 IOPS]
    Random Write 4KiB (Q= 1,T= 1) : 72.412 MB/s [ 17678.7 IOPS]
     

  5. janos666

    janos666 Master Guru

    Messages:
    970
    Likes Received:
    142
    GPU:
    MSI RTX2060 6Gb
    I recall the real hit was usually observed at <4k block IO (below pagesize) or at 4k with really fast devices (think >=10G ethernet, PCI-E 4x NVMe SSD, etc).
    It would be interesting to see these 3(+1) scenarios benchmarked:
    pre-S MC + OS-SM off
    post-S MC + OS-SM off
    post-S MC + OS-SM on
    pre-S MC + OS-SM on (this fourth combo doesn't make much practical sense to me:, although it could show how much the software-only mitigations [if any exist - ?retpoline?] slow the system down).
    However, I personally in favor of using the latest MC (regardless of security) and leave the mitigations at default (for security) anyways, so I don't really have an incentive in doing these tests. And any benchmarks are dubious since it's well knows these mitigations affect certain workloads, so their effect on general overall system performance can vary a lot (depending on what you do with your PC). A single benchmark score will hide or magnify the effects by orders of magnitudes at times.

    I guess any MC-only mitigations have little to no effect on performance (in case OS options are disabled for the main mitigation features, since the rest usually depends on those) and you can turn all (main) options off in the registry (it's well documented, not a secret, not an unsupported tweak), except may be retpoline (I am not sure if that's always active, even for completely unaffected CPUs or different code-paths exist).
     
    Last edited: Apr 10, 2020
    RodroG and Smough like this.
  6. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    The performance hit is way smaller than I would have expected, still, I am keeping all of this stuff off. I have my reasons and I firmly believe this isn't needed for the average home user, if you know what you are doing on your computer, I very much doubt you can receive an "attack".
     
  7. EdKiefer

    EdKiefer Ancient Guru

    Messages:
    2,517
    Likes Received:
    241
    GPU:
    MSI 970 Gaming 4G
    It has never been a big hit I ever saw, CPU BM or anything I can measure is even smaller.
     
    Smough likes this.
  8. BlindBison

    BlindBison Master Guru

    Messages:
    382
    Likes Received:
    61
    GPU:
    RTX 2080 Super
    So, uh -- is all of this still an issue at this point in 2020? Should I be utilizing Wagnersoft's program or has Microsoft cleared this up by now?

    Seems like a really bizarre oversight on their part to begin with if I'm honest so I'd hope they would've cleared it all up by now, but it's hard to tell since game stuttering can come from a variety of sources as I understand it.
     
  9. aufkrawall2

    aufkrawall2 Master Guru

    Messages:
    673
    Likes Received:
    80
    GPU:
    RX 480 8GB OC/UV
    BlindBison likes this.
  10. BlindBison

    BlindBison Master Guru

    Messages:
    382
    Likes Received:
    61
    GPU:
    RTX 2080 Super
    Thanks!
     

  11. Braska

    Braska New Member

    Messages:
    3
    Likes Received:
    0
    GPU:
    EVGA GeForce GTX
    So what are best funktion to use here, are there any completed data here or is it all preferences?
     
  12. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    Just install Windows 1809 and fully update it, best Windows version imo ATM and its 100% free of the Standby memory list bug that needs to be "cleaned" or else games will stutter. 1709 and early 1803 versions have this problem, which I is why I wouldn't ever recommend any of them. Too old, many bugs that don't help in gaming.

    1909 doesn't have this problem either, obviously, so if you are on that one, its all good.
     
    Xtreme512 likes this.
  13. Xtreme512

    Xtreme512 Master Guru

    Messages:
    678
    Likes Received:
    9
    GPU:
    GTX1060 6GB
    what about forcing timer resolution to 0.5 for better fps in games putting heavy usage on HT (crysis 3)? is this ever fixed?
     
    BlindBison likes this.
  14. disq

    disq Member Guru

    Messages:
    103
    Likes Received:
    55
    GPU:
    GTX
    I'm on 1909 and i still have to use ISLC to clear the standby list as it takes all the free memory

    Cleaning it one time is enough thought (sometimes (few) two are necessary) contrary to previous versions of Windows where it needed to get cleaned several times

    Windows x64 with 16GB of RAM
     
    Last edited: May 3, 2020
  15. aufkrawall2

    aufkrawall2 Master Guru

    Messages:
    673
    Likes Received:
    80
    GPU:
    RX 480 8GB OC/UV
    Are you aware that this itself is not and never was an issue?
     

  16. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    It doesn't need to be cleaned if your games are not stuttering lol. That's the point, to have it filled while you play a game so it caches stuff from your RAM lmao. This is NOT an issue, this is how Windows is mean to work, the bug was Windows NOT CLEANING standby memory, the bug was game stutter while it was full, because this is the way games operate, except the game doesn't expect the O.S to be bugged and cause a mess.

    Do you clean it because your games still stutter or for no reason? If you do it for no reason, may as well stay on 1709, where is needed because it has the bug.
     
    Xtreme512 likes this.
  17. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    I just force it to 1ms and I also use bcdedit /set useplatformtick yes so the timer is "even". I guess 0.5 still helps at some games but never noticed a problem.
     
  18. disq

    disq Member Guru

    Messages:
    103
    Likes Received:
    55
    GPU:
    GTX
    Yes, guess i forgot to state the obvious, the games stutter as soon as there's no free memory available, sometimes i even forget to launch ISLC and then while gaming i notice the stutter, open ISLC and can see that there's no free memory available
     
  19. Smough

    Smough Master Guru

    Messages:
    528
    Likes Received:
    89
    GPU:
    GTX 1060 3GB
    Then your problem lies elsewhere, maybe you need to update a driver or its bugged. The issue was solved by both nvidia and MS. Old nvidia drivers had this problem, along with O.S issues from old Windows builds. Also some games will crash or will become a stutter fest if you clean the standby memory constantly. This method was a band-aid, its not a fix.
     
  20. BlackNova92

    BlackNova92 Active Member

    Messages:
    64
    Likes Received:
    7
    GPU:
    16gb
    that's actually good to know tbh.
    i was constantly using it to clear my sbl and for the 0.5ms
    i'm on 1903 at the moment, you think i should do what you said and just use the timer and leave the sbl alone for now?
    i only play overwatch at the moment and haven't had any stutter issues so far.

    how can i actually disable it from cleaning it?
    by using zeroes on both ends?
     

Share This Page