1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Another look at HPET High Precision Event Timer

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Bukkake, Sep 18, 2012.

  1. khanmein

    khanmein Ancient Guru

    Messages:
    1,649
    Likes Received:
    71
    GPU:
    EVGA GTX 1070 SC
    My result. Any comment?

    [​IMG]
     
  2. Koniakki

    Koniakki Ancient Guru

    Messages:
    2,836
    Likes Received:
    436
    GPU:
    ZOTAC GTX 1080Ti FE
  3. mbk1969

    mbk1969 Ancient Guru

    Messages:
    7,495
    Likes Received:
    4,035
    GPU:
    GeForce GTX 1070
    Nice looking program for true nerds/geeks.
     
  4. khanmein

    khanmein Ancient Guru

    Messages:
    1,649
    Likes Received:
    71
    GPU:
    EVGA GTX 1070 SC
    Seriously? o_O

    After I ran the TimerBench, I went to check on the Event Viewer and noticed a lot of error regarding TimerBenchHookService64.
     

  5. mbk1969

    mbk1969 Ancient Guru

    Messages:
    7,495
    Likes Received:
    4,035
    GPU:
    GeForce GTX 1070
    Report to developer.
    (I promised to not comment anymore on HPET BS, and I keep my promise.)
     
  6. khanmein

    khanmein Ancient Guru

    Messages:
    1,649
    Likes Received:
    71
    GPU:
    EVGA GTX 1070 SC
    Okay, I personally feel is totally a gimmick and out-dated tool.
     
  7. mbk1969

    mbk1969 Ancient Guru

    Messages:
    7,495
    Likes Received:
    4,035
    GPU:
    GeForce GTX 1070
    I was not trolling or sarcasming. If you want to be a responsible user report events you mentioned to developer.

    Sure program has no fancy animated stuff, but it comes with built-in test, with quite detailed stats, with proper explanations and with clean non confusing layout.
     
  8. alexander1986

    alexander1986 Member Guru

    Messages:
    167
    Likes Received:
    16
    GPU:
    RTX 2060
    this timerbench app, tried messing around a bit with things like disabledynamictick yes/no , forcing resolution to 0.5 ms, platformclock yes/no and reboot before testing app again, got similar results all the time, only setting useplatformclock to yes gave me noticeably worse results :S


    [​IMG]


    Also, just wondering, what is the difference if any, between having useplatformclock set to no in bcdedit or deleting the value completely? my HPET is disabled in bios btw, thanks !
     
  9. mbk1969

    mbk1969 Ancient Guru

    Messages:
    7,495
    Likes Received:
    4,035
    GPU:
    GeForce GTX 1070
    This question was answered in this thread several times.
     
    alexander1986 likes this.
  10. alexander1986

    alexander1986 Member Guru

    Messages:
    167
    Likes Received:
    16
    GPU:
    RTX 2060
    I see, my bad :p will check through the pages then !
     
    insp1re2600 and mbk1969 like this.

  11. alexander1986

    alexander1986 Member Guru

    Messages:
    167
    Likes Received:
    16
    GPU:
    RTX 2060
    @mbk1969 yo, another question about timers, this timer resolution tool that sets it to 0.5 ms for example instead of the default 1ms many games seem to request, including fortnite that I play the most, any valid reason to use it over the 1 ms timer ? and any possible negatives like increased load on cpu or input lag or something even if higher fps?

    Ive tested it off and on but in this game on my setup cant really say it makes a difference atleast not a big one, so asking more in theory I guess :p

    cheers !
     
  12. mbk1969

    mbk1969 Ancient Guru

    Messages:
    7,495
    Likes Received:
    4,035
    GPU:
    GeForce GTX 1070

    Windows supports such objects as timers. Developer can use timer to fire notification (one time or repeatedly) on expiration of timeout. That timeout value is set in milliseconds. If developer sets timeout smaller than the system timer resolution the timer will fire at desired timeout only occasionally because kernel fires (expires) all existing timers only when system timer fires (special timer interrupt). For example if developer sets 1ms timeout for the timer but system timer resolution sits on default 15.6ms the developer`s timer will fire at 15.6ms most of the times. But Windows offers the function to set desired system timer resolution, and developer can set it to improve timers behaviour in his code.

    If developers of Fortnite request 1ms system timer resolution you can`t improve their code by requesting 0.5ms instead (unless you think they are stupid).

    In Windows 7 setting the system timer resolution to smaller milliseconds increase the load on all CPUs/cores because in Windows 7 all CPUs/cores get the timer interrupt.
    But starting with Windows 8 the OS kernel becomes tickless and only one of CPUs/cores get the timer interrupt. So in Win8+ you will not increase the CPU load that much.
    And of course modern CPUs should sustain good performance even with system timer resolution 0.5ms, but if you have some badly coded driver or application CPU can be overwhelmed.

    PS With default 15.6ms system timer resolution the timer interrupt fires 64 times per second. You can calculate the differences between 15.6ms and 1.s and 0.5ms load.
     
    Last edited: Mar 9, 2019
    alexander1986 likes this.
  13. alexander1986

    alexander1986 Member Guru

    Messages:
    167
    Likes Received:
    16
    GPU:
    RTX 2060

    I see, always good information in your answers! thanks a lot and cheers for now :)
     
  14. dexterav

    dexterav Active Member

    Messages:
    66
    Likes Received:
    10
    GPU:
    Evga GTX1080Ti FE
    about HPET, if i enabled this crap my NVME ssd performance is only 1/3 of normal, thats easy decide ON or OFF
    in fact is visible slowest than normal
     
  15. tsunami231

    tsunami231 Ancient Guru

    Messages:
    9,447
    Likes Received:
    311
    GPU:
    EVGA 1070Ti Black
    is this still a thing? people still play with this? unless you have horrid latency issue you shouldnt be messing with this imo
     
    vf likes this.

  16. dexterav

    dexterav Active Member

    Messages:
    66
    Likes Received:
    10
    GPU:
    Evga GTX1080Ti FE
    i dont play with this, just once try benchmark, must have HPET on
    an after this i discovered this abysmal performance with HPET on, i have this allways off
     
  17. tsunami231

    tsunami231 Ancient Guru

    Messages:
    9,447
    Likes Received:
    311
    GPU:
    EVGA 1070Ti Black
    OS should be using what is best for said system, a majority of times it does, some times it dont, which only time is should be messed with. atlest from what I understand on how it works and my tests on multiple systems, I could be wrong im sure @mbk1969 understand it much better then me.
     
  18. mbk1969

    mbk1969 Ancient Guru

    Messages:
    7,495
    Likes Received:
    4,035
    GPU:
    GeForce GTX 1070
    @tsunami231

    Thank you for credits. dexterav has not specified how exactly he turned HPET off - in BIOS or just in BCD. Also he has tested his rig and then he has claimed something. So let him be.
     
  19. dexterav

    dexterav Active Member

    Messages:
    66
    Likes Received:
    10
    GPU:
    Evga GTX1080Ti FE
    idont have options about HPET in bios
    old x99 asus ws have opt hpet on/off in bios z370f not
    its my experiences with HPET, i dont say its happend everyone who enble HPET, its PC everything possible :D
     
    Last edited: Mar 10, 2019
  20. lime

    lime Member

    Messages:
    22
    Likes Received:
    1
    GPU:
    1070
    I don't know what all this discussion is about. TimerBench should be tested with forced HPET on and then another test with HPET off (default)
    Even if its enabled in the BIOS, or like on ASUS mainboards no option at all, it is still disabled until needed.
    So then you compare HPET on (with bcdedit commands) with off and you can clearly see which one is better on your rig.
     
    dexterav likes this.

Share This Page