[SOLVED] Windows XP system clock losing (lots) of time.

[4wd]

I wonder if anyone has any ideas with the problem below before I just go ahead and reinstall XP.

I've got a headless XP Pro system that runs the following programs 24/7:

uTorrent - BitTorrent
XAMPP   - Webserver
SABnzbd - Usenet
DNTV Scheduler - DVB-T recorder

It runs no AV and only the basic Windows Firewall, access is via RDP - no programs are executed on it apart from NexusFile for a bit of file management.  I've checked it over using MBAM - nil found.

It's done this for months on end without a problem, (close to two years), but in the last couple of weeks or so Windows has been losing time to the tune of minutes per hour, (eg. just in the last couple of hours, 18 minutes).

I now get it to sync to au.pool.ntp.org every ten minutes, in that time it can lose 2 minutes.

Currently I'm using ClockMon which gives the RTC and system time along with any difference between them - I've set this to check every minute and sync to the RTC if the difference is greater than 100ms.

Task Manager shows the system as being idle pretty much of the time, with the CPU, (AMD 235e), barely hitting 20% at minimum clock, (800MHz), even when capturing two DVB-T streams.

I guess the question is: Anybody got any ideas what kind of process could be stopping interrupts or does anyone know how I can find out?

[40hz]

Check the internal CMOS battery. Most likely it's either dead or dying. That's usually what causes that problem.  

[justice]

I agree

[Ath]

It's real hardware, and not a VMWare image running on a virtual host? 'Cause VMWare 'clients' can lose time if the VMWare tools stuff isn't installed or updated when the host is updated to a newer version. If it's not installed, you'd probably lose several hours per day in time though

Battery is most likely cause if it's hardware, like 4wd said.

[Stoic Joker]

'nother vote for the CMOS battery. I've run into this one a few times myself.

[4wd]

I should have mentioned I've changed the battery but the fact that the RTC doesn't lose time at all and that Windows system time is correct at boot and starts losing time from that point would imply that the RTC isn't the problem.  If it was I would expect to see the time incorrect from the moment the system boots.

Also, from all that I've read, Windows doesn't use the RTC once it's booted - it's then down to system interrupts.

Just another note: I have two of these systems, with the same hardware, (MB, CPU, RAM, case + PSU) - one runs XP Pro the other Win7HP x64.  I swapped the batteries between them - the W7 system runs fine and the XP system still loses time.  The problem didn't move with the battery.

@Ath: Yes, real hardware.  ZOTAC Mini-ITX Geforce 8200 iGFX, Athlon II 235e & 4GB DDR2-800 - see it here.

EDIT: Just had a thought, when I wake up tomorrow, (err, later today), I'll boot it from a PE disc and see what happens.  But now, I really should be asleep  

[Carol Haynes]

What time server is being used on your system - try changing to an alternative. I had a problem once where the time was set correctly and then suddenly the time changed and it was a TS causing the problem.

http://support.microsoft.com/kb/314054

Also check your time zone of set correctly (Control Panel - International).

Last thought: if not the CMOS battery have you checked the stability of your PSU?

[Stoic Joker]

Have you tried running

Also, from all that I've read, Windows doesn't use the RTC once it's booted - it's then down to system interrupts.

-4wd (December 19, 2011, 07:58 AM)

True, have you tried running w32tm /reset to force is to recalculate the offset?

[4wd]

What time server is being used on your system - try changing to an alternative. I had a problem once where the time was set correctly and then suddenly the time changed and it was a TS causing the problem.

-Carol Haynes (December 19, 2011, 08:26 AM)

I now get it to sync to au.pool.ntp.org every ten minutes, in that time it can lose 2 minutes.

-4wd (December 19, 2011, 06:30 AM)

So it uses a random NTP server every time.  I have used pool.ntp.org, (even more random), but I have also just turned the resyncing off - it still starts losing time from the moment it's booted.

Also check your time zone of set correctly (Control Panel - International).

One of the first things I checked.

Last thought: if not the CMOS battery have you checked the stability of your PSU?

Thanks, that made me think of something of something else, a long shot.....but still a shot.

True, have you tried running w32tm /reset to force is to recalculate the offset?

-Stoic Joker (December 19, 2011, 08:55 AM)

I don't seem to have the /reset switch on w32tm, /resync might be what you mean?


Crap!  I think I found the problem    At some point I must have disabled or forgotten to enable the HPET and WDRT in BIOS when playing with the battery.

I've just reset the BIOS and the RTC/OS difference is back to being under a second every 5 minutes.

Thanks for the suggestions, I think I need to go back to tech school  

Maybe I can just chalk it up to a "Senior Moment" now I've turned 50.  

BTW, can anyone point me to something that says what the WDRT actually does, (apart from being a watchdog)?

Addendum: Spoke too soon, instead of steadily losing minutes per hour like before, it now appears that the XP system clock can lose hours overnight.  This is usually accompanied by the inability to access the system via RDP the next day - this is starting to look more like some software has gone haywire.  I guess it's down to a OS reinstall....oh well, didn't have anything planned today.

[4wd]

Well, decided to switch to shotgun methodology and changed the battery again, (even though the RTC doesn't lose time), removed a PWM fan controller I installed about the time it went nuts and restore the system to approx. 2 months prior.

Back to thumb-twiddling to see what happens.

Computers....who'd have 'em?

[f0dder]

Hm, would a CMOS battery affect the RTC of a running system? I WouldaThunk it only used the battery when powered off?

[4wd]

Hm, would a CMOS battery affect the RTC of a running system? I WouldaThunk it only used the battery when powered off?

-f0dder (December 25, 2011, 09:04 AM)

That's what I said in my previous posts, RTC time has no effect on Windows OS once it's up and running, (unless you're telling Windows to get it from the RTC, of course, like I am with ClockMon).

Windows uses interrupts once it's running but for the sake of completeness I changed it yet again, even though I'd already proved it was OK.

Currently the system is behaving itself so AFAIAC it comes down to one of two things:
1) Software that I installed post-October, or
2) The PWM fan controller generating sufficient interference to upset the clocking.

I have had the second happen once before but on the 240V mains system, a slave/master switch inducing enough interference into the mains circuit to cause a new digital clock on the other side of the house to run fast....returned the clock twice before I worked out what it was 

[4wd]

Found it!.......well almost.

Something locking up the networking on the machine which seems to stop:
1) the clock updating
2) any kind on RDP, VNC access
3) disk access

which necessitates me having to push the power button.  However, the various programs serving web pages still pump out pages when accessed but just fail to interact.

I'm starting to think I have a clash between the various servers running and the ports they're using.  Time for a bit of rationalisation and a completely fresh reinstall.

[cranioscopical]

Found it!.......well almost.

-4wd (January 07, 2012, 12:31 AM)

 

Tenacious 'R' Us 

[worstje]

I did some checking, and it looks like it may be driver related. Especially the chipset. Any chance Windows Update decided it knew best when you weren't looking?

[4wd]

I did some checking, and it looks like it may be driver related. Especially the chipset. Any chance Windows Update decided it knew best when you weren't looking?

-worstje (January 07, 2012, 07:57 AM)

I don't have WU set to Auto on this machine, normally just Notify only.  But of course that doesn't preclude me from clicking OK in a burst of stupidity.

And as far as nVidia chipset drivers go, nothing would surprise me if that turns out to be the problem - one of the reasons I try to avoid their chipsets now. (Addendum: As if to prove a point their website won't even give me the XP drivers, it keeps sending the Vista/7 ones.)

Thanks for that, I'll download the latest from nVidia then remove the current ones with DriverCleaner and see what happens.  If that doesn't work then I'll go back to my previous regimen of nlitening the install and never updating past SP3.  It worked that way for years without a problem.

Tenacious 'R' Us  

-cranioscopical (January 07, 2012, 07:50 AM)

I'd never make any money doing repair work - I'd never know when to just say 'Screw it!' and give up.

[4wd]

I did some checking, and it looks like it may be driver related. Especially the chipset.

-worstje (January 07, 2012, 07:57 AM)

Guess what?  You hit it in one, after an image restore I removed the nVidia driver and then reinstalled the latest WHQL XP drivers I could find, (which wasn't from nVidia' site).  No more clock problems since.

Definitely no more updating for this machine now, service disabled.

Thanks for the clue worstje  

[worstje]

I'm glad you figured it out!

I also want to express my thanks to Cranioscopical for his community spirit; little surprises like the one he sent my way are what makes this community great!

[4wd]

I also want to express my thanks to Cranioscopical for his community spirit; little surprises like the one he sent my way are what makes this community great!

-worstje (January 23, 2012, 05:11 AM)

+2 

(Senior moments and all)