From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S264595AbUEDTEv (ORCPT ); Tue, 4 May 2004 15:04:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S264597AbUEDTEv (ORCPT ); Tue, 4 May 2004 15:04:51 -0400 Received: from grex.cyberspace.org ([216.93.104.34]:53510 "HELO grex.cyberspace.org") by vger.kernel.org with SMTP id S264595AbUEDTEt (ORCPT ); Tue, 4 May 2004 15:04:49 -0400 From: Hal Nine Message-Id: <200405041905.PAA05084@grex.cyberspace.org> Subject: Clock skew with hyperthreading and high load? To: linux-kernel@VGER.KERNEL.ORG Date: Tue, 4 May 2004 15:05:24 -0400 (EDT) Cc: hal9@cyberspace.org MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org I have a Intel(R) Xeon(TM) CPU 2.40GHz (stepping 9) running under 2.4.24 with hyperthreading enabled. I have noticed that, under extremely high loads (high swap usage including, but this may not be relevant) the system clock looses a lot of seconds. For example, when running ntpdate against a reliable time source every hour, on some runs it corrects only fractions of seconds, but on other runs it ajusts up to 200 seconds (during exactly those periods of high load). Is this a known problem? Any good solutions/workarounds? h