ClockBlocker (profiled AMD power-management control)

Discussion in 'Videocards - AMD Radeon Drivers Section' started by A2Razor, Dec 10, 2015.

  1. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    If possible, the next time that you see something like this happen -- assuming that ClockBlocker's main process is still working properly. I'd like for you both to do three things for me, this would be a great help in figuring out exactly what's happening. (so that I can narrow this down and get it fixed)


    -Check the color of the tray-icon, if it's reverted to yellow or still red.
    -tray-icon right-click > Misc > Dump to TXT > "Log" and "Settings" (run them both and save the resulting files somewhere)
    -Open a Task-Manager and look for if you see any helper modules running.


    If you can attach those files (or put the contents in a code block in a post) and let me know the result of the tray-icon, etc, that'll be a huge help in figuring out at least what part of ClockBlocker this is happening in. Also fine to PM them to me on the forums, but in a reply to the thread is fine too.

    --Also, when you close ClockBlocker to restart the program, do you have to manually terminate a helper-container? Or do you just need to close it from the tray-icon and re-launch ClockBlocker? (this will hint if the helper process is completely hosed or not)


    EDIT: One last question: Did you have anything happen at the same time as the game-crash indicated in the Windows Event Viewer (system events) such as a TDR or other driver recovery? -Basically so that I know if this is a typical program crash or something more.

    And thanks! That should hopefully be enough to get me started.
     
    Last edited: Jan 29, 2016
  2. JonasBeckman

    JonasBeckman Ancient Guru

    Messages:
    17,292
    Likes Received:
    2,628
    GPU:
    MSI 6800 "Vanilla"
    Tray icon is red, that I did check the last time it happened, it's also how I understood it was still blocking the downclock function.

    Can't get the log data now but I can do that next time something like this occurs, will try to remember to do that.
    (AMD CCC / CNext has Overdrive on but unmodified besides the power draw threshold or what it was called again which is at 20%, pretty sure I could set it to 50% and it would be totally safe but eh I'm used to when it capped at 20 on older GPU models. :D )

    (For Afterburner I do have hardware monitoring on but no voltage or clock adjustments though a manual fan curve has been defined as this GPU has a tendency to go for noise over cooling though even at default speeds of 20 - 30% it's still providing really good GPU cooling, I just increased it a tad up to 50% instead and then curve it upwards to full fan speed as temps exceed 70 degrees celsius.)

    As for exiting and restarting clock blocker I just right clicked on it in the tray and shut it down normally, started it again and the icon was yellow, again the next time it happens I'll try to see if the helper process is also running and if it's still running after Clockblocker has been shut down.

    No driver crashes or other issues, I was downsampling which stresses the GPU a bit so I noticed when it crashed because with AMD's VSR at least if you downsample and the shutdown isn't clean it tends to re-arrange the current open desktop windows / folders as if it was fitting them to the downsampled resolution so every time a game does crash it's a bit of a hassle to resize them all back again. :p
    (Doesn't take too much time though and for some games this happens regardless, don't know what exactly causes this behavior but I'm guessing it has something to do with how the process - or well game - is shutting down.)
     
    Last edited: Jan 29, 2016
  3. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    That would imply that either ClockBlocker still thinks the process is running and matching a rule (potentially the process handle being stuck open), or the process scanner thread being hung (stuck). However, if the scanner was hung, the main UI thread would wait on this on closing ClockBlocker -- eg, it'd never close properly. As ClockBlocker's main thread holds open a Mutex for its duplicate instance checking, you'd know immediately if another instance was running because the next launch wouldn't start due to the duplicate-check.


    --So, I'm going to assume that it's closing properly, which means it's unlikely that this is a total stall. (which is good) but...
    On your rules, it sounds like most of your rules are exclusions and that you're using 3D-Program detection, so this is a foreground trigger on the process right? -- This contradicts my first assertion that the scanner thread isn't stalled, because even if ClockBlocker assumed that the program was still running and even if the scanner still detected it as present, it'd no longer be the foreground process. (which "should" be excluding the 3D-Program detection and hints more in the opposite direction that the scanner thread is hung in some way)

    I'll probably wait until I see the log, yet this is definitely helpful. Since the tray-icon is red and the helper container is closing properly on exit, it's very likely that this is a problem in the main program (ClockBlocker's controller process).
     
  4. Fukki

    Fukki New Member

    Messages:
    6
    Likes Received:
    0
    GPU:
    MSI R9 390

    Didnt had any crashes or shutdown problems, shutdown was normal and when I turned PC on again a few hours later, they were stucked at 1100/1550 (MAX I use), even so, game performance was really bad like this (diablo 3 had massive frame drops, barely playable), when I turned overdrive ON I had to press the reset button for everything go back to normal.

    overdriver on is not causing any problems for now.
     

  5. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    -Even if I can't reproduce this, I'll see if I can add in some detection for such a state just on the description and 'fix' it by reverting to the stock state on startup.

    Thanks again for letting me know!
     
  6. mR Yellow

    mR Yellow Ancient Guru

    Messages:
    1,935
    Likes Received:
    0
    GPU:
    Sapphire R9 Fury
    The issue with the clock speed locking at stock speeds was fixed by toggling official and unofficial mode. Note, i use the Compute method.
     
  7. LocoDiceGR

    LocoDiceGR Ancient Guru

    Messages:
    2,289
    Likes Received:
    785
    GPU:
    Gigabyte 3060 Ti
    I had a crash yesterday, (on the division beta) using clockblocker on default settings BUT I search it, and today i saw a post on reddit about the same problem and the same card,

    Black screen, No signal and then nothing...maybe its a driver crash and no recover, thats why the black screen!

    Also the guy said he crashed on Rainbow6!

    I think clockblocker is fine, but the AMD driver needs fix, (as usual)!

    --(Bad english sorry, not my first language!) :)
     
    Last edited: Jan 30, 2016
  8. fat4l

    fat4l Member

    Messages:
    19
    Likes Received:
    0
    GPU:
    Asus Ares 3 8GB
    ok ok guys, I'm getting confused now :D

    Which of the 4 methods is the best one to use ?
    3rd ?

    When I used 1,2,4 with my crossfire setup, but crossfire was disabled in drivers, it bumped my clocks up on the disabled card as well, which is not wanted obviously.

    Does the 3rd method put the less load on the card itself than the first 2? I mean, is there any performance drop?

    Thanks !!! :)
     
  9. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    Currently that's one of my theories [on the clock-states too, not just blackscreen (eg, failed TDR) ], that something happened (such as a display driver crash) that caused AMD's drivers to restore the default (non overclocked) clocks. Compute-method doesn't touch clock-states, so it's most likely that something else triggered a revert. --Toggling between unofficial and official mode would cause the overclock to be re-applied in Afterburner.

    You might try disallowing Crimson's components (such as the CP) from starting with the machine and to see if this has any effect, but I am not sure that it will. People used to stop CCC from starting with the machine to get OC's from Afterburner to stick when starting their PC rather than using a startup-delay in AB (until after CCC started).


    If the third method works for you, it has the least performance impact. But, this doesn't mean necessarily best as the compute and ADL methods work completely different (while achieving the same goal -- high clockspeed).

    #3 (ADL-method standalone, and also #4) adjusts graphics card performance levels, eliminating the lower clock-states as choices.
    #[1 & 2], & 4 keep a compute kernel active for each graphics device detected, which in turn cause the drivers to select a higher clock-state.


    While there is a small performance cost to the compute-method (first [the default] and second choices), keep in mind that the compute-method does not touch any settings. That means there's less chance for the machine to get in to some funky state through things like driver-crashes. It can also be argued safer in that lower clock states are not eliminated as choices.
     
    Last edited: Jan 31, 2016
  10. LocoDiceGR

    LocoDiceGR Ancient Guru

    Messages:
    2,289
    Likes Received:
    785
    GPU:
    Gigabyte 3060 Ti
    I dont have AB installed, and i have no OC on my card.

    Overdrive is disabled of course, i have only installed Trixx for Custom fan speed.
     

  11. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    Yeah, I was kindof replying to both of you at once, because I was borrowing and chaining thoughts on your comment with the TDR's and blackscreens. I know that you weren't necessarily running Afterburner and or were running at stock speeds, so that was more at mR Yellow on the clock speeds being reduced and reset with AB.

    I may have misunderstood though too, did you have your clockspeed get 'stuck' or drop, or reduce after this crash in the division? Or just a blackscreen crash and lockup?


    --Just making sure, so I can try to connect the dots with everyone's unique setups and what they see.

    Thanks again!
     
  12. marshy61

    marshy61 Member

    Messages:
    12
    Likes Received:
    0
    GPU:
    HD7970M
    Hey
    Thanks to this program i dont have downclock problem anymore but when i try to overclock with evga precision x ,after 10 15min my gpu turn back to unoverclock speed (1000/1400 to 850/..) and i have to restart my pc to overclock again.Is there any fix to this problem?

    btw cant overclock with msi ab.
     
    Last edited: Jan 31, 2016
  13. Anvi

    Anvi Member Guru

    Messages:
    128
    Likes Received:
    0
    GPU:
    R9 390 8GB
    A2Razor, Thank you so much!

    This has fixed a lot of stability and corruption issues I've had on R9 390 card!

    If AMD keeps broken powerplay with Polaris, then I hope this tool still works :)
     
  14. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    We're talking about an AMD card here right? --I wasn't aware that Precision X worked on AMD (both Alexey's code before the separation and EVGA's replacement), so I'll have to take another look at this. Admittedly I've not looked at EVGA Precision in a very long time.

    I'd thought that EVGA was exclusive to selling NVidia graphics cards and only concerned with their products.

    You're certainly welcome, and I'm hopeful at this point due to Intel, NVidia, and AMD hardware working on the compute method, that at least this method will keep working. We shall see!

    -Sadly, I can't guarantee that I will be buying a Polaris to test with, unless AMD fixes Fiji's display corruption officially (or maybe some company is so kind as to provide one). But, you never know everything may keep working with no changes too.
     
    Last edited: Feb 1, 2016
  15. marshy61

    marshy61 Member

    Messages:
    12
    Likes Received:
    0
    GPU:
    HD7970M
    Yeah its amd 7970m and only evga 4.2.1 version is working
     

  16. Tetus

    Tetus Member

    Messages:
    14
    Likes Received:
    0
    GPU:
    MSI R9 290x Twin Fozr
    Well apparently the clock blocker does not work very well with AC Syndicate, sometimes the clock oscillates, any clue to fix this ?
     
  17. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    Tried the ADL and compute method standalone? (1st and 3rd choices, and or combined -- 4th)
    Are you overclocking the card, and if so with which tool(s)? What type of clock behavior are we talking here? (if you can describe that oscillation a bit more -- frequency, the clocks you're seeing, etc)


    AC Syndicate is another game that I've never played and don't own, but if you can give me an idea of the behavior maybe there's something I can do.

    --I'm assuming this is on your desktop and not the laptop with the Intel IGP, but as much as you can give on the OS, driver version, etc, may help too.
     
    Last edited: Feb 3, 2016
  18. Tetus

    Tetus Member

    Messages:
    14
    Likes Received:
    0
    GPU:
    MSI R9 290x Twin Fozr
    I'm using the compute method and the custom clocks of my desktop's GPU (MSI 290x)

    The clockblock detects the game just fine and starts to do it's job then after a while a playing the clock starts to decrease as much as 100MHz for some periods, and when this oscillation starts it doesnt stop anymore, it keeps happening and happening.

    Some of my specs:
    Windows 10
    Crimson 16.1.1 hotfix (this issue was happening with 16.1 as well)
    GPU: MSI r9 290x
    CPU: I5 4690k

    EDIT: Tried every method available, still no sucess AC Syndicate doesnt let clock block work properly
     
    Last edited: Feb 3, 2016
  19. fat4l

    fat4l Member

    Messages:
    19
    Likes Received:
    0
    GPU:
    Asus Ares 3 8GB
    Hi.
    I think I found a bug with Crossfire AMD.
    I have 290X in crossfire(Asus Ares 3 card) and I'm running aida64 all the time therefore I made a new rule to keep the clocks down(2D) as far as aida64.exe is running and I'm using "ADL-force downclock" method for 2D.
    Then I have a rule for my game that I play(World of Warcraft for example) to keep the clocks up(Block if in foreground method) and I set it to priority 1 so it overrides the rule set for aida64.
    When running the game, all is fine, clocks are stable(3D).

    However, if i go to windows and the game is in the foreground, the clocks should go into 2D mode.
    My #card1 clocks are forced to be 2D but on my #card2, my clocks are still 3D.

    This is happening with "ADL block-if in foregroud" method for 3D(world of warcraft) and "ADL-force downclock" method for 2D(aida64).

    I think that ADL doesn't apply to a second card properly(when running crossfire) when set to " force-downclock" or basically, the second card doesn't "see" that the game is in the backround(the rule for "block only if in the foreground) and that it should apply the "ADL-force downlock" rule.

    If I turn the game off, the clocks are set correctly, 2D, for both cards.

    [​IMG]


    :)
    Any help ?


    Edit://
    forgot to say I have to use ADL methods.

    Edit2://
    also sometimes the second card's clocks are fluctuating.
    Again, maybe a problem with crossfire(second gpu)?

    Edit3://
    Ok and now, after the restart, my gpu1 clocks are stuck in 2D mode. Even W/o Clockblocker running, my gpu1 freqs are stuck @2D. I think I have to reinstall the driver now ..:3eyes:
     
    Last edited: Feb 3, 2016
  20. A2Razor

    A2Razor Master Guru

    Messages:
    519
    Likes Received:
    85
    GPU:
    6800XT, XFX Merc319
    @ fat4l:

    Yes, it does sound like you've found an issue with Crossfire (per the inability to force downclocks, possibly also with the fluctuation). I don't have any Crossfire setup to test with though, so I'll have to just hand inspect everything and see if I can find anything.

    Managed to reproduce this finally. What's happening is whenever a person defines rules that cause ClockBlocker to be applying a clock-state when the machine is shutdown, Windows issues a request for ClockBlocker to close and then does not "wait" sufficiently for the clockspeeds to be reverted. ClockBlocker starts closing down, attempts to revert the clock-states, which fail as the machine is too far along in the shutdown process. This results in the machine getting in to a bad state on the next start.


    Will be fixing the issue by refusing the shutdown until the clockspeeds are correctly reverted and also running checks on startup if we get in to a bad state like these.

    --No ETA on this just yet [I've been rather busy lately], yet I don't imagine it'll take me that long once I get some time.


    And yes, you'll either need to clean drivers and reinstall, or alternately to enable Crimson CP's overdrive and hit revert. For now: You can prevent this from happening by closing ClockBlocker manually before shutting down or by closing AIDA (causing all your rules to stop being evaluated, and thus for ClockBlocker to revert the prior clock-states).


    *To be clear to everyone else-- what I've managed to reproduce only effects ADL use, and it only effects you if you have rules that cause adjusted clock-states to be active while Windows is shut down.

    ===========================
    To put it another way:

    Blocking rules for games. => Probably safe because you'd typically close a game before shutdown.

    However, let's say that the game crashes (black screen) and you cannot get the game to close... Then you forcedly shutdown Windows (in an effort to close the game). => This might get you in to a bad state.
     
    Last edited: Feb 4, 2016

Share This Page