From owner-freebsd-acpi@FreeBSD.ORG Wed Dec 24 02:01:54 2008 Return-Path: Delivered-To: freebsd-acpi@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 072051065675 for ; Wed, 24 Dec 2008 02:01:54 +0000 (UTC) (envelope-from smithi@nimnet.asn.au) Received: from sola.nimnet.asn.au (paqi.nimnet.asn.au [220.233.188.227]) by mx1.freebsd.org (Postfix) with ESMTP id 629188FC1F for ; Wed, 24 Dec 2008 02:01:53 +0000 (UTC) (envelope-from smithi@nimnet.asn.au) Received: from localhost (localhost [127.0.0.1]) by sola.nimnet.asn.au (8.14.2/8.14.2) with ESMTP id mBO21p6s089821; Wed, 24 Dec 2008 13:01:51 +1100 (EST) (envelope-from smithi@nimnet.asn.au) Date: Wed, 24 Dec 2008 13:01:50 +1100 (EST) From: Ian Smith To: Dmitry Kolosov In-Reply-To: <200812231941.11998.ivakras1@gmail.com> Message-ID: <20081224122702.C29108@sola.nimnet.asn.au> References: <20081221233822.7E92545020@ptavv.es.net> <200812221927.00568.ivakras1@gmail.com> <20081223140958.I29108@sola.nimnet.asn.au> <200812231941.11998.ivakras1@gmail.com> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Cc: freebsd-acpi@freebsd.org Subject: Re: Problem on AMD64 X-BeenThere: freebsd-acpi@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: ACPI and power management development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 24 Dec 2008 02:01:54 -0000 On Tue, 23 Dec 2008, Dmitry Kolosov wrote: > On Tuesday 23 December 2008 06:40:50 Ian Smith wrote: > > On Mon, 22 Dec 2008, Dmitry Kolosov wrote: > > [..] > > > > > Could you give to us some links about powersaving with EST? For now, i'm > > > using powerd: > > > powerd_enable="YES" > > > powerd_flags="-a maximum -b adaptive -n adaptive -r 30 -i 35" > > > in my rc.conf. I'm not on AMD64, so i'm sorry, powerd works well to me > > > (125 MHz on battery and 2.16GHz on AC), BUT battery life time is equal > > > in both cases and something about 50 minutes, so i think powerd is not > > > so powerfull for me. > > > > -i percent Specifies the CPU idle percent level when adaptive mode > > should begin to degrade performance to save power. The > > default is 90% or higher. > > > > -r percent Specifies the CPU idle percent level where adaptive mode > > should consider the CPU running and increase performance. > > The default is 65% or lower. > > > > I don't think your powerd running and idle percentages are likely to > > work too well; too close together, and too close to the 'busy' end. > > > > Try stopping powerd (/etc/rc.d/powerd stop) then running powerd manually > > in verbose mode in its own console (powerd -v [flags]) to watch how it > > behaves under varying loads. > > > > I suspect that you will find it 'flapping' between some frequencies too > > often at constant load, as there's insufficient hysteresis between the > > idle/running marks. Compare it with using the default -i and -r and if > > those aren't suitable, try rather smaller variations from the defaults. > > > > If it lacks responsiveness, try decreasing the polling interval. > > > > cheers, Ian > > Thanks Ian, > -i and -r values was selected (3 month ago) by long playing with `powerd -v > [flags]` in foreground, as you advice to me. The behavior of powerd is correct, > and parameters was selected correctly to my environment. Also, > debug.cpufreq.lowest was set to 500, and it starts to increase with 30% of CPU > load, and decrease (to 500) with 35%. It's just good to me and it perfectly > works. Fine if it works for you. However your settings are 30% idle (70% busy) and 35% idle (65% busy), not percent loaded as you've expressed it here. > I'm realy think it's all about my hdd (WD), its toooooo hot, even when no file > activity, even when lid is closed, on battery or on AC, no matter. It's about > 60-65C. I agree, 60C is too hot for a HDD. I'd get nervous over ~50C myself. Have you tried cleaning out the dust from the air passages and fan/s? > How to decrease the polling interval? -p ival Specifies a different polling interval (in milliseconds) for AC line state and system idle levels. The default is 500 ms. With a shorter interval, say 250ms, you may find that a lower value of debug.cpufreq.lowest will be responsive enough, which might save power. cheers, Ian