[ntp:questions] LAPIC timer and PIT

Kevin Oberman oberman at es.net
Tue Sep 4 15:45:49 UTC 2007


> From: "Jason Rabel" <jason at extremeoverclocking.com>
> Date: Tue, 4 Sep 2007 09:29:48 -0500
> Sender: questions-bounces+oberman=es.net at lists.ntp.org
> 
> 
> > Spoon wrote:
> > 
> > > On a typical x86 motherboard, are the PIT (Programmable Interval Timer) 
> > > and LAPIC (Local Advanced Programmable Interrupt Controller) timer 
> > > driven by the same crystal?
> > > 
> > > In other words, will ntpd compute a similar frequency offset if I switch
> 
> > > from PIT to LAPIC timer?
> > > 
> > > Is the LAPIC timer in any way "better" than the PIT (resolution? 
> > > precision? accuracy? overhead? anything else?)
> > 
> > Would someone care to comment?
> > 
> > Also, I'd be happy with links or pointers :-)
> 
> You have only one (or *maybe* two) crystal(s) on your MB. A PLL chip takes
> that and generates all the various frequencies necessary for the components
> to operate on the board. Stability & quality all falls back on that (cheap)
> crystal. It doesn't matter how good the timer is if your timebase is no
> bueno.
> 
> IIRC there is documentation on enabling / disabling APIC in the NTP Wiki.
> 
> As the name implies, APIC is a controller, not a timer. Don't get the two
> confused.

I think that the common confusion of APIC and ACPI may be causing
this. ACPI does provide high-quality clocking in most newer systems. 

As Jason pointed out, all of the clocks are usually derived from a
single oscillator, but the mechanism of dealing with them can make a
great deal of difference on short-term accuracy. Simply stated, some
timers are processed in ways that make them a bit more stable. ACPI-Fast
is generally about the best choice for precision clocking on systems
where ACPI is fairly recent and works correctly.

Since you don't state what OS you are using, I can't say much about how
to select the clock.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 224 bytes
Desc: not available
URL: <http://lists.ntp.org/pipermail/questions/attachments/20070904/eff4b569/attachment.pgp>


More information about the questions mailing list