Another look at HPET High Precision Event Timer

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

  1. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    Not to me. Tiny fluctuation, also it can be that the recording program is affected itself (either along with mouse polling or alone).
    I always was sceptical about 1000 Hz mouse polling. Imo, it is a marketing stuff, because I doubt average human can use mouse so fast that he will benefit from 1000 Hz polling.
    Only full scale blind tests can prove me something.
     
  2. aufkrawall2

    aufkrawall2 Ancient Guru

    Messages:
    4,352
    Likes Received:
    1,815
    GPU:
    7800 XT Hellhound
    It's 1000Hz to ~guarantee that no display refresh is missed, as these clocks aren't synced.
     
  3. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    500 Hz guarantees that too. And 250 Hz should too, even for 244 Hz displays.

    PS But my doubts are about human capabilities - can they change mouse position every 1 ms to do something meaningful?
     
  4. aufkrawall2

    aufkrawall2 Ancient Guru

    Messages:
    4,352
    Likes Received:
    1,815
    GPU:
    7800 XT Hellhound
    I don't think that 250Hz would be enough for 244Hz, as afair 125Hz produce a stuttery mess even with 75Hz display refresh rate.
    Anyhow, why not keep it practical: 1000Hz are no issue for any half-decent PC from the last ten years.

    Though I don't know why anyone would deliberately make themselves unhappy by looking at weird synthetic stats. If you don't notice mouse jumping when doing some aim tests in games or by moving some windows, there likely isn't any potential left for noticeable optimizations.
     
    mbk1969 likes this.

  5. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    Unneeded load. If decent rig will have badly coded device driver then such unneeded load can lead to practical glitches like stutters.
    Also unneeded power usage when you just read or watch something even not touching the mouse.

    Especially when you do not know how synthetic tests were implemented and what exactly they show.
     
  6. PanosXidis

    PanosXidis Member Guru

    Messages:
    149
    Likes Received:
    11
    GPU:
    3080 SUPRIM X
    disabledynamicktick and useplatformtick is way placebo on 2020 seriously guys on cpuz with disabledynamicktick yes i have worst perfomance than default so you dont need nothing
     
  7. tsunami231

    tsunami231 Ancient Guru

    Messages:
    14,725
    Likes Received:
    1,854
    GPU:
    EVGA 1070Ti Black
    i am surpised people still mess with this window10 should be using the best timer dependent on the system at this point in time and has been doing so for while. only reason to mess with it is if you have same system latency issue, even then if you have them it probably a driver issue.

    I been leave it alone all this time and have no issue latency wise
     
  8. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    @tsunami231

    Agree.
    Several latest comments were about system timer resolution and mouse poll frequency, though.
     
  9. Th3Awak3n1ng

    Th3Awak3n1ng Guest

    Messages:
    6
    Likes Received:
    0
    GPU:
    GTX 980 4GB
    It's not placebo. I tested it on my system:

    useplatformtick yes = sets timer resolution 0.500 instead of 0.499 but gives 1200-1300 DPC spikes, so I have noticeable stutters during playing.

    useplatformclock true = significantly improves mouse response in game, DPC spikes to 200-300 while playing the game, everything is smooth, but graphics performance is 5-10% worse.
     
  10. theahae

    theahae Active Member

    Messages:
    67
    Likes Received:
    12
    GPU:
    GTX 1060
    you feel difference with the mouse because it became worse, not better.
     

  11. Th3Awak3n1ng

    Th3Awak3n1ng Guest

    Messages:
    6
    Likes Received:
    0
    GPU:
    GTX 980 4GB
    I see you have troubles with reading, I will make it clear for you:

    default (no timer tweaks) = bad mouse input
    bcdedit useplatformclock true = good mouse input
    bcdedit /deletevalue useplatformclock = bad mouse input
     
  12. shm0

    shm0 Master Guru

    Messages:
    231
    Likes Received:
    51
    GPU:
    6800XT
    Isn't the windows kernel tickless since win 8 anyway ?
    So does changing the timer frequency actually do anything?
    I guess only when disabledynamicktick is set to yes?
    But high kernel tick also means high overhead ?
     
  13. janos666

    janos666 Ancient Guru

    Messages:
    1,648
    Likes Received:
    405
    GPU:
    MSI RTX3080 10Gb
    I think current Windows uses "dynamic idle ticks", not "fully dynamic ticks". But even then, NO_HZ_FULL on Linux "[...] try to shutdown the tick whenever possible", a rate will still be configurable (it's the maximum rate, I guess) and a single CPU will continue running the ticks.
     
  14. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    https://www.kernel.org/doc/Documentation/timers/NO_HZ.txt

    For second mode - OMIT SCHEDULING-CLOCK TICKS FOR IDLE CPUs - only idle CPUs do not get periodic ticks.

    For third mode - OMIT SCHEDULING-CLOCK TICKS FOR CPUs WITH ONLY ONE RUNNABLE TASK:
    So at least one CPU (core) will always get periodic ticks.

    I honestly doubt that on consumer PC with up to 32 cores there ever will be a situation when each core executes only one task (thread). In Windows there are more than hundred processes with more than thousand threads - by default, so to speak.
     
    Last edited: Oct 29, 2020
  15. MakeHate

    MakeHate Guest

    Messages:
    10
    Likes Received:
    1
    GPU:
    MSI 1660ti
    [Written using a translator]
    Faced with a problem that I can't solve, I ask for help or advice:

    Characteristics:
    Ryzen 2700, 4.0 GHz OC.
    MSI B450 A-Pro Max, the latest BIOS.
    1660ti Ventus OC.
    Goodram Iridium 3333MHz CL15 OC, 2x8GB.

    The problem is with timers. By default, HPET is disabled and everything is fine, however, I found a problem that in CPU-Z data for all timers does not converge and there is a "stutter" when testing these timers every 3-4 seconds for a fraction of a second. ACPI and PCI converge, but RTC lags behind them sometimes from 0.12 to 0.36 milliseconds. As far as I know, they should go perfectly straight. I've tried a lot, dozens of different commands and actions related to bcdedit /enum and the like, including the debugging modes tscsyncpolicy Default / Legacy and Enchaned.

    Can you help me?
     
    fluidz likes this.

  16. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    @MakeHate

    Why do you think that some timers information shown by CPU-Z means issues for real apps like games?
    Do you play games? Do games work with issues?
     
    MakeHate likes this.
  17. MakeHate

    MakeHate Guest

    Messages:
    10
    Likes Received:
    1
    GPU:
    MSI 1660ti
    [Written using a translator, don't get mad]

    Noticed problems in Lara Croft last. Cutters with the sound disappearing for a couple of seconds, which categorically does not agree with the data from Latency Moon and DPC Latency Checker, which claim that I have absolutely no delays. And besides, I'm worried :(
     
  18. MakeHate

    MakeHate Guest

    Messages:
    10
    Likes Received:
    1
    GPU:
    MSI 1660ti
    delete this, sorry
     
  19. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
    @MakeHate

    Do not worry. It can be that RTC is not used in latest Win10 versions.

    DPC Latency Checker is not compatible with Windows 10.
     
    MakeHate likes this.
  20. MakeHate

    MakeHate Guest

    Messages:
    10
    Likes Received:
    1
    GPU:
    MSI 1660ti
    You are, I understand, the most experienced in this business. I'm using your improved standard timer. Another Intelligent standby list cleaner claims that I have 0.4991 ms delay now with your timer. Is this normal?

    Thank you very much for everything.
    I'm not one of them, by the way. I also saw one guy with the same problem here, and there was another person with a similar problem on the Russian-language Internet.
     

Share This Page