[CentOS] cpuspeed problems with 5.2 and 2.6.18-92.1.18 kernelrevision
jses27 at gmail.com
Thu Dec 18 17:24:32 UTC 2008
> -----Original Message-----
> From: centos-bounces at centos.org
> [mailto:centos-bounces at centos.org] On Behalf Of Kai Schaetzl
> Sent: Thursday, December 18, 2008 7:18 AM
> To: centos at centos.org
> Subject: Re: [CentOS] cpuspeed problems with 5.2 and
> 2.6.18-92.1.18 kernelrevision
> ArcosCom Linux User wrote on Thu, 18 Dec 2008 00:00:29 +0100 (CET):
> > The last try:
> > 1) Power off the laptop along 2/3 hours.
> > 2) Power on and log into gnome.
> > 3) The cpuspeed monitor shows that all is working fine.
> I don't know what this monitor is. There's a cpuspeed service
> script. If
> enabled it will load the necessary kernel modules and the configured
> governor. Config is in /etc/sysconfig/cpuspeed. You can also
> check the
> script yourself to see what it does. It may very well be that
> it thinks that
> your CPU has a flaw that stops it from correctly working with
> scaling. I know that the script has checks for AMD CPUs in it
> for instance.
> If there is some cpuspeed GUI monitor, this has nothing to do
> with the
> cpuspeed service script. If cpuspeed is disabled the
> performance governor
> should be on. As I see you know where to check this. So, as I
> said earlier,
> shut it off (chkconfig cpuspeed off), reboot, and check the values.
> Also, if you think that this has something to do with Gnome
> then you should
> *not* boot in the GUI and see if you get better results.
> If that doesn't help there's more likely a problem with the
> driver. Have you
> checked that it is loaded?
> > # cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
> > 800000
> > (and I think this last value is the problem because I can't
> change it with
> > echo "1733000" >
> I tend to agree. That's why you want to check this value
> without gnomem
> without cpuspeed and also right after booting up (as you say
> scales down after a few minutes, so theoretically there
> should be a higher
> value in the beginning and there might indeed be a problem in
> the driver
> that stops it from scaling up again). It's also possible that
> the algorithm
> for calculating the speedup is doing something wrong, so it
> doesn't scale up
> again, because it thinks it's not necessary. You could set
> the UP_THRESHOLD=
> to something very low like 5
> > 4) Some minutes after, the speed goes down to 800 MHz
> and no more wants
> > to grow up. I can't change to force performance or using
> userspace and
> > select the speed.
> > Any help more about?
> > If I launch any heavy compilation (as kernel compilation),
> the speed don't
> > want to grow up. Really is working fine under these
> circumstances with the
> > performance governor selected?
> No, but this wasn't obvious from your earlier posting. As I
> said: did you
> stop cpuspeed? This is not a xen kernel, isn't it?
> About the top posting/qoting etc.: The point is to make your
> postings as
> readable for others as possible. If you quote what you really
> answer to and
> then answer that and then quote the next part you answer and
> answer that
> it's much easier to follow for the reader and also much
> easier for *you* to
> answer as you cannot easily overlook questions if you go thru
> one by one.
> And everything that you don't answer (including the signature
> etc.) simply
> doesn't belong in the quote and is omitted. You can take this
> posting as an
Kai is right about the cpuspeed deamon. If you stopped the service it should
and will run full speed. But there is a CATCH "dynamic cpu speed features".
This is in your BIOS Settings thay need to be disabled. Most all Mother
Boards that Supports Intel Centrino has it. Check that out.
More information about the CentOS