[ntp:questions] micro-optimization

Martin Burnicki martin.burnicki at meinberg.de
Thu Mar 12 08:59:00 UTC 2009


Augustine wrote:
> On Mar 11, 4:19 am, Terje Mathisen <"terje.mathisen at tmsw.no">
> wrote:
>>
>> I do know that Intel intends to make processor-based timers more useful
>> by setting up a counter which always runs at a fixed rate, independent
>> of any frequency power stepping. On such a cpu RDTSC would again be the
>> best choice, but hopefully the OS would realize this and start to use it
>> for QPC.
> 
> What both Intel and AMD did on Core2 and Phenom was to move the "TSC"
> over to the BU.  Therefore, RDTSC latency grew to typically 60 to 100
> cycles with, and this is worse for NTP, jitter, because of the
> different clock domains involved.

That's interesting.

Do you have some pointers where this is explained?

Martin
-- 
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany




More information about the questions mailing list