Content-Type: multipart/mixed; boundary="===============2771927170559609561==" MIME-Version: 1.0 From: Kristen Carlson Accardi Subject: Re: [Powertop] Tunables: Ondemand Governor and intel_pstate Date: Mon, 23 Sep 2013 12:10:12 -0700 Message-ID: <20130923121012.4a237c32@bluebird.jf.intel.com> In-Reply-To: 20130920075259.GA6749@mobster.intranet.ufz.de To: powertop@lists.01.org List-ID: --===============2771927170559609561== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Fri, 20 Sep 2013 09:52:59 +0200 Christian Krause wrote: > Dear PowerTop Developers, > = > I have been wondering for a while and just found out: > = > tunables show: > = > Bad Using 'ondemand' cpufreq governor > = > but: I don't have a ondemand cpufreq governor -- that is what I just foun= d out: since some kernel > versions ago the intel_pstate driver is the default for my cpu: > = > # zgrep PSTATE /proc/config.gz > CONFIG_X86_INTEL_PSTATE=3Dy > = > # cat /sys/devices/system/cpu/cpu[0-9]*/cpufreq/scaling_driver > intel_pstate > intel_pstate > intel_pstate > intel_pstate > = > intel_pstate does not provide ondemand governor, just powersave and perfo= rmance. It saves power by > putting the cpu in sleep mode instead of decreasing the frequency, see id= le stats: C7-SNB 98.7% ... > = > Please include a check in PowerTop so ondemand governor does not get sugg= ested for intel_pstate. > That was really confusing at first to have powertop suggest that but me n= ot having ondemand > governor. > = > = yes, I've been meaning to take that suggestion out of the tunables all together for a while, as the value of using ondemand is pretty questionable. --===============2771927170559609561==--