System clock running very fast

[MK-2]

this seems a weird one to me.

just got in the new job and sorting things out in the office. i noticed that on the pc i use, the clock was out by aabout ten mins, so i altered the time to match my phone exactly.

a few hours later, and the pc time is ten mins ahead. if you open up the clock and watch the hand moving around, it appears to be faster than a normal clock.

 

i've looked on google but a lot of the options relate to bittorrent clients and things like that which i know aren't installed on here.

i also dont have the 'internet time' tab to sync it an NTP server.

 

it's obviously not a major issue as i still have my phone and a clock on the wall, but i'm just thinking if it becomes an issue school wide

[powdarrmonkey]

What time does your friendly local DC think it is?

[Domino]

Do your machines not all look at a server to get their time?

 

Ours all get it from the DC which in turn gets it from a NTP server online.

 

The time needs to be fairly accurate for Kerberos.

[Michael_84]

As in the previous post the time is synced with the domain controller (the first one created in the forest I believe) so when you change it locally on your system it will just sync back with the domain controller at some point.

 

So the way to get around this is either keep setting the clock on the domain controller to match the correct time or simply use NTP this is configured on the domain controller by using:

net time /setsntp:<>

 

I recommend you use pool ntp servers, also make sure you have NTP open on your firewall otherwise it won't sync.

 

Once you've set the NTP server on the domain controller you can resync the time by doing w32tm /resync

 

Good luck

[SYNACK]

I have had this kind of issue before on a single machine but much more extreme, it thought a full minute had passed each second but other than that the system seemed fine.

 

It turned out to be an issue with the network adapter as when I replaced it the problem went away. It may be possible that an add on device in your system is also generating power spikes on the pci bus causing extra clock ticks.

 

Additionally you could also check the BIOS battery in the system to make sure that it is still providing the right voltage, it should not effect the clock but is still worth checking.

[somabc]

You can follow this guide

 

Spiceworks Community - Microsoft Authoritative Time Servers

[MK-2]

ok ive set the server time to the correct one, resynced mine, and tomorrow ill have a look on isa and check that ntp is allowed.

 

**edit**

within the time of me writing this, the server (and therefore my machine) have gone 1 minute faster than my phone.

I'll leave this til tomorrow as to be honest I have bigger things i need to work on.

thanks for the help though everyone, i'll come back to this tomorrow

[OverWorked]

It's probably the CMOS battery running down. I lower voltage can make the crystal oscillator run a bit faster.

 

The lower voltage means it doesn't have as far to switch from the top to the bottom of the cycle, making each cycle a little bit shorter than it should be.

[srochford]

It's probably the CMOS battery running down. I lower voltage can make the crystal oscillator run a bit faster.

 

The lower voltage means it doesn't have as far to switch from the top to the bottom of the cycle, making each cycle a little bit shorter than it should be.

 

errrr?? Are you sure???

 

As I understand it, the quartz crystal has a resonant frequency which should be very stable. It does depend on a sort of feedback loop and I'd guess if the battery is running down then that might be affected but I wouldn't have thought that that mattered when the computer was running from mains power.

 

It could be a faulty capacitor - back in the olden days (:-)) digital clocks used to have a variable capacitor so you could fractionally adjust the timekeeping. Don't think they are ever fitted to PC motherboards - I'm guessing that they're supposed to be more stable and not need adjusting but if a cap goes leaky or something then it might fail.

 

You can configure how often the time service checks against the server; if you set this to (say) every 5 minutes then that ought to keep it stable.

[OverWorked]

errrr?? Are you sure???

 

I know it sounds far fetched, and I was expecting such a response, but I've seen it first hand.

 

Way back when I was doing my Electronic eng degree, I had a crystal oscilator on my final year project running fast. My tutor told me I was underpowering it. He was right (he should have been - he had a PhD in electronic eng). I increased the voltage to the what it should have been and it kept the frequency it was supposed to.

 

I've seen several old PCs do the same. The clock starts running fast a few weeks or months before the CMOS battery goes.

[Domino]

 

It could be a faulty capacitor

 

I read that as flux capacitor.....

 

 

I didn't think servers could reach 88 miles an hour (or 1.21 gigawatts for that matter)

[OverWorked]

I need to make full retraction of what I've said in this thread.

 

Last night I thought long and hard about what I believed of crystal oscillators. I now recall that the problem I had with the oscillator in my electronics project, and the conversation I had with my electronics tutor, was about low input voltage distorting the output waveform, not the frequency.

 

It was a false memory of an event almost 20 years ago. For some reason I've been carrying around the belief of crystal oscillators speeding up under low voltage for all this time without questioning it. It's a good job I never had a career in electronics.

 

As for the CMOS clocks on old PCs with low batteries slowly gaining time: I must have imagined it, based on that false memory.

 

It's funny how the mind works. I've been alone all week, stuck in a empty school, and my mind's going. I need a holiday.

 

Sorry!

 

 

However, it is possible that the crystal oscillator on your mobo is just bad (it does happen). See these results. In which case, there's nothing you can do about it, except replace the mobo.

Edited August 7, 2008 by OverWorked

[SOSAGES]

i have had 2 pc's (one hp server 1 dell box) have issues with time.

 

flashing the BIOS would always fix the problem for a day or so but then it would come back. Replacing the CMOS battery fixed the problem and both boxes are running fine today.

[OverWorked]

Replacing the CMOS battery fixed the problem and both boxes are running fine today.

 

Ha! So I hadn't imagined it. A low battery can affect the RTC.

 

Thanks SOSAGES, I thought I was going mad.

[SOSAGES]

unless we are both mad..:hat:

 

but it was the first thing HP server support suggested when i called.

[OverWorked]

We can't rule that out.

 

I'm sure that in the past I've found old PCs that have been stuck on shelves for a long period unused with RTCs showing dates well into the future - months or years. Whenever I've seen it I've always thought "Uh Oh... battery's going".

 

When the battery dies altogether, the RTC is likely to default back to a date in the past, usually the date the BIOS was written, so they'll have, say, 1988 for the date. They wouldn't default to some arbitary date in 2010 for example. Finding an ancient PC with a RTC date in 2010 has always meant to me 'low battery'.

 

I've always believed, rightly or wrongly, that when a battery drops its voltage, the RTC can speed up. If HP support agree then that's good enough for me.

 

Can I retract my retraction? I was riddled with self-doubt earlier, but have been bolstered by SOSAGES.

[MK-2]

sorry i didn't get around to replying. i set the server time yesterday, synched this pc and it seems to have kept it. cant believe it was something as simple as the server time being off.

 

so much going on around here that i'm forgetting the basics!!

 

thank you to everyone for your help

[powdarrmonkey]

The Kerberos time has a tolerance of about 5 minutes I think, so if the client and the server are out the client will catch up, but without jumping (as this will invalidate your current Kerberos tickets). That's why you see it moving quicker than normal.

[MK-2]

aha, that sounds like what it was doing

[SYNACK]

Ha! So I hadn't imagined it. A low battery can affect the RTC.

 

It would depend on how badly the output waveform is distorted, as the system will probably count high to low edges if the waveform is distorted enough it could be sufficient to trip the count logic more often than it is supposed to. The crystal could also be damaged as you say, a crack in the crystal will change its output frequency. I remember reading about people modding old crystal oscillators by opening them and chipping off small amounts of the crystal to speed up the clock effectively overclocking the device. This would be stupidly hard now as the crystals are much smaller, especially for pc bus speeds.

[powdarrmonkey]

You can see how far out you are with:

 

w32tm /stripchart /computer:

 

where somedc is a valid domain controller or other NTP source.

[srochford]

Sorry!

 

 

no problem - everyone makes mistakes but I'm impressed that you did think it through and remembered all the details. It certainly makes sense that the waveform will get distorted (listen to a battery powered amp when the batteries are going flat) and if something needs nice sharp edges then it will have problems if it doesn't get them.

[ahuxham]

net time \\DC-NAME-01 /set /y

 

In your login script, syncs your clock with the DC's clock.

[srochford]

May be better to use it in the machine startup script - by default users don't have the right to change the system time.

 

Also, the "net time" command is deprecated - in common with all the other old Lan Manager commands they may not be supported in the future. You can replace it with w32tm /resync