From owner-freebsd-hackers Tue Dec 26 18:31:59 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id SAA27439 for hackers-outgoing; Tue, 26 Dec 1995 18:31:59 -0800 (PST) Received: from zygaena.com (zygaena.com [206.148.80.1]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id SAA27434 for ; Tue, 26 Dec 1995 18:31:50 -0800 (PST) Received: (from nobody@localhost) by zygaena.com (8.6.11/8.6.9) id VAA17589 for ; Tue, 26 Dec 1995 21:32:06 -0500 From: Will Brown Received: from lochsa.i.com(198.30.169.3) by zygaena.com via smap (V1.3) id sma017587; Tue Dec 26 21:31:52 1995 Received: by lochsa.i.com (Smail3.1.29.1 #4) id m0tUleP-000VUKC; Tue, 26 Dec 95 21:31 EST Message-Id: Date: Tue, 26 Dec 95 21:31 EST To: freebsd-hackers@freebsd.org Subject: clock accuracy? Sender: owner-hackers@freebsd.org Precedence: bulk Why does the clock on my DTK (ASUS?) PCI P100 running FreeBSD 2.0.5 gain about 32s per day? At this rate I need to run ntpdate about every 20 minutes to avoid clock steps (ugh). I found various postings from various folks on topics such as the use of freq. synthesizers instead of crystals on newer PeeCee boards, and the woeful state of the venerable 8254. Is that the basic issue? Do these PeeCee's really not have a timekeeper that is even as good as my $25 Timex? If so, is there a software workaround (eg: hack some timing parameter in the kernel somewhere?). What about the stability of the primary timing source on these newer pentium boards? Are they likely to maintain a consistent clock drift? Can I tweak some parameter based on the measured drift and expect it to stabilize things? I also seem to recall seeing some mention of clock drift due to console messages (or was I hallucinating?). Sorry if this is a rehash - ptrs to answers appreciated! -- Will Brown (ewb@zygaena.com)