I time drift 'backward' on reboot would seem to imply that your box
is 'running fast', and a reboot reverts to the hardware TOD clock on
reboot?
I haven't noticed gross time errors on my quad xeon, but I do sync the
time twice a day using 'rdate' to a machine that is our local ntp sync
box (a solaris host with only one cpu and a pretty accurate set of time
sources on the net).
I have heard others complain about the drift, but I haven't seen it, yet
my machine often is 400% compute bound and at other times is screaming on
3 10Krpm U2W SCSI drives at the same time.
Btw, interactive response still seems quite good even when smashing
the machine...
Robert Hyatt Computer and Information Sciences
[EMAIL PROTECTED] University of Alabama at Birmingham
(205) 934-2213 115A Campbell Hall, UAB Station
(205) 934-5473 FAX Birmingham, AL 35294-1170
On Thu, 29 Apr 1999, Ingo Molnar wrote:
>
> On Thu, 29 Apr 1999, Ralf Baechle wrote:
>
> > > But this could not be a solution, I think. Why is this systemclock
> > > drifting anyway?
> >
> > The systemclock, that is the jiffies counter, is interrupt driven. If
> > the system misses an interrupt, you're loosing 1/HZ, that is 0.01s for
> > an Intel system. In particular SCSI systems tend to loose interrupts.
> > The clock jitter caused by this effect is often so bad that xntpd looses
> > synchronisation.
>
> i'm afraid this does not explain my time drift. I occasionally see a time
> drift between reboots as well (it could be a jump of 5 or 10 minutes
> backwards?), but it's rather sporadic. I reboot very often :) (and i've
> been running 'missed tick' detectors for some time and my box is not
> losing timer IRQs). So something else has to be the reason ... I'm not
> using xntpd.
>
> -- mingo
>
>
> -
> Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
> To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
>
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
