[ntp:questions] Influence of temperature on PC oscillators

Uwe Klein uwe_klein_habertwedt at t-online.de
Sun Apr 15 08:08:05 UTC 2007


Hal Murray wrote:
> In article <4620bcdf$0$7511$426a74cc at news.free.fr>,
>  Spoon <devnull at invalid> writes:
> 
>>Hello,
>>
>>I've found the following articles rather interesting:
>>
>>http://www.wraith.sf.ca.us/ntp/#bios-issues
>>http://www.ijs.si/time/temp-compensation/
>>http://en.wikipedia.org/wiki/Voltage-controlled_oscillator
>>
>>Regards.
> 
> 
> I don't agree with the suggestion to disable the spread spectrum
> stuff.  It's there for a reason.  Effectively, your CPU will run
> slightly slower.  I haven't seen anything published to indicate
> that the speed is sufficiently less stable so that ntpd will notice.
> It will be different than the nominal value, but that's not a problem.

The issues are enumerated here:
	http://www.adriansrojakpot.com/Speed_Demonz/BIOS_Guide/BIOS_Guide_02d.htm#Spread Spectrum

This chip forex:
	http://www.digchip.com/datasheets/parts/datasheet/115/W40S01-04.php
does its stuff by slowly sweeping Fosz up and down.
(keep in mind that with modern MoBos the clocking is done by vco/pll and a
rather low xtal ref )

if ntpd stuff runs in near sync with this sweep you get _very_ funny effects.
( ntpd may start to see the mixing results of the ss sweep and its own
rep rate )

> 
> It would be fun to see if you can measure it.
The nVidia nForce2 issue i posted some time ago is more pronounced with
spread spectrum enabled ( indicating a timing issue) .

uwe




More information about the questions mailing list