Windows time runs fast after sleep

Joined
Jun 8, 2016
Messages
16
I've been running a 7900X and Asus B650E (MicroCenter bundle) for a few months now. With Expo set in bios everything was fine until last week when Windows started blue screening with corrupt kernel errors. Then the PC wouldn't even post, only in safe mode. The only way to get it running stably since then was to set everything to manual and just not change any settings. So now my memory is running at 4800 instead of the 6000 it's supposedly rated for. On cold boots everything is fine. But after waking from sleep the clock in Windows runs fast. The longer the machine is on the further off it gets; I've seen it 9 or 10 minutes fast before Windows re-syncs.

What might cause this?
 
No ideas on what might be the issue here? I've just been shutting down completely every day rather than using sleep mode and everything is fine that way. Super weird.
 
Sleep mode has been an issue for overclockers (sometimes those who just run XMP, DOCP, Expo) long enough that I stopped using it quite a few years ago. I've not bothered to try it again for years, and with how fast PCs now boot with SSDs, I'll likely never bother.
 
Sleep mode has been an issue for overclockers (sometimes those who just run XMP, DOCP, Expo) long enough that I stopped using it quite a few years ago. I've not bothered to try it again for years, and with how fast PCs now boot with SSDs, I'll likely never bother.
Oddly enough sleep mode actually works properly when running Expo with no other changes. It's only when I change the Core Ratio that I start having problems. And even that was working for months before the motherboard decided to stop posting with those settings. Very odd behavior.
 
It's probably that whatever timer windows is using for the clock is changing frequency during sleep. Why that happens, I dunno, and good luck figuring it out. If I run into clock abnormalities, the go to is turn it all the way off and back on and it usually clears up. Hardware/software is great like that.

NTP sync can 'discipline' small clock abnormalities, but typically gives up when you're off by several minuted a day.
 
It's probably that whatever timer windows is using for the clock is changing frequency during sleep. Why that happens, I dunno, and good luck figuring it out. If I run into clock abnormalities, the go to is turn it all the way off and back on and it usually clears up. Hardware/software is great like that.

NTP sync can 'discipline' small clock abnormalities, but typically gives up when you're off by several minuted a day.

To be clear the clock is actually still accurate when it wakes from sleep. It's just that seconds start ticking faster at that point and after ~10 minutes it's a full minute ahead. At any rate, I've posted this a few places and haven't even seen anyone complaining of the same issue so it looks like sleep is just off the menu for as long as I have this build. Bummer considering how long this platform takes to boot, but c'est la vie.
 
To be clear the clock is actually still accurate when it wakes from sleep. It's just that seconds start ticking faster at that point and after ~10 minutes it's a full minute ahead. At any rate, I've posted this a few places and haven't even seen anyone complaining of the same issue so it looks like sleep is just off the menu for as long as I have this build. Bummer considering how long this platform takes to boot, but c'est la vie.

Yeah, I get that. I mean that the timer windows for timekeeping uses was running at frequency A before sleeping, and at frequency B after sleeping. Timekeeping during sleep might be totally different (might use the RTC for example). On FreeBSD or Linux you can easily change the timer used for timekeeping and that might help with something like this. I can't find anything about that for Windows; just a lot of info about the time synchronization service. There are some references that suggest Windows will use the TSC if it's present and marked invariant; so my guess is the TSC is present, marked invariant, and comes up at a different frequency after sleeping, at least sometimes.

Microsoft does have a github repo called Windows-Time-Calibration-Tools which might be helpful, but probably a lot of work to narrow down the problem, and then not much you can do about it after. I'd wonder if the tsc frequency is modulated by that Core Ratio setting you mentioned, and if sometimes your system comes back from sleep with that setting at defaults; that might be easier to validate than anything else.

Edit: might be worth trying to force HPET timing?
 
Last edited:
net stop w32time
w32tm /unregister
w32tm /register
net start w32time
w32tm /resync

reboot

possible malware too
 
net stop w32time
w32tm /unregister
w32tm /register
net start w32time
w32tm /resync

reboot

possible malware too

Thanks for the suggestion but I tried this a few weeks ago and it made no difference unfortunately. Malware sounds unlikely. It is (was when this started) a completely fresh install of Windows. Also it doesn't happen when Core Ratio is left unchanged (or Expo settings used).
 
Back
Top