Vcore rising during "idle" and lowering during load.

Discussion in 'Die-hard Overclocking & Case Modifications' started by PhugoidEffect, Jan 28, 2018.

  1. PhugoidEffect

    PhugoidEffect Guest

    Messages:
    3
    Likes Received:
    0
    GPU:
    1080TI
    Hello folks!
    The behaviour described in the subject is happening in my recent build of a 8700K running at 5 GHz in a X Hero. I remember to have disabled all energy savings stuff and set all cores to synchronise at 50, I've also set a LLC of 4 (very middle of the range). How can I prevent the voltage from rising at idle? During load-idle it ranges about from 1.29 to 1.35.
    Thanks in advance.
     
  2. RealNC

    RealNC Ancient Guru

    Messages:
    4,959
    Likes Received:
    3,235
    GPU:
    4070 Ti Super
    This is normal and as it should be. Voltage regulators aren't infinitely accurate. To protect the CPU from overvoltage, the voltage is slightly lowered when more power is needed by the CPU (happens during load.)

    Some mainboards have a setting that disables this protection. It's usually called "vdroop" (yes, two "o"s). If you disable it (or lower it), the vcore difference between load and idle is smaller, but your CPU is getting overvoltaged for a brief period of time.

    More info here:

    https://www.masterslair.com/vdroop-and-load-line-calibration-is-vdroop-really-bad

    However, another contributor in vcore drop is the PSU. In that case, the issue is called "vdrop" (one "o"). This is not fixable and requires a new PSU that has better voltage regulators.
     
  3. user1

    user1 Ancient Guru

    Messages:
    2,748
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    the voltage dropping under load is normal behaviour

    consider more agressive llc for more stable voltage under load, and turning on C1E to reduce voltage at idle ,

    becareful with setting the llc, it can actually set the voltage to be higher than what you set in the bios under load if set too high.
     
  4. PhugoidEffect

    PhugoidEffect Guest

    Messages:
    3
    Likes Received:
    0
    GPU:
    1080TI
    Thanks for the clarification!
    I'm still tuning the overclock and now it seems to be stable at Vcore 1.290 and LLC 6, with much better temps under load (and in "idle" as well). This PC was built for single task only and will not run 24/7 (although occasionally might be used for like 48 hours in a row), and as such it doesn't matter to me the idle state (I even disabled everything related to savings, C-States, etc).
    I've been trying voltages from 1.260 (in 0.010 steps) and all of them up to 1.290 showed Windows Hardware Errors in the HWiNFO64 sensor section, 1.290 seems to prevent this, and the top temperatures running the stress test of AIDA64 are topping at 75C. Vcore reported by HWiNFO64 ranges from 1.280 to 1.296 (idle and load conditions together, which seems strange to me, very close, but probably due to LLC 6).
     

  5. PhugoidEffect

    PhugoidEffect Guest

    Messages:
    3
    Likes Received:
    0
    GPU:
    1080TI
    Just an update: the software I use (a flight simulator) is crashing (.DLL reported by event viewer) at a certain point, which is not common at all, so I lifted a little more and now running 1.300 (and always reported as 1.296 by HWiNFO64, either idle or load, no changes).
     
  6. RealNC

    RealNC Ancient Guru

    Messages:
    4,959
    Likes Received:
    3,235
    GPU:
    4070 Ti Super
    Overclocking without running a stress test to see if it's stable, is not recommended. Errors can result in things like garbage being written to disk, for example, which can cause file corruption, file system damage, etc.

    Applications crashing is the visible effect of instability. Silent data corruption you don't see until it's too late is the invisible part of it.
     

Share This Page