high cpu temp and fan speed problem

Wed May 12 18:24:43 UTC 2010

2010/5/12 Ian Smith <smithi at nimnet.asn.au>

> On Mon, 10 May 2010, Nathan BIAGINI wrote:
>  > 2010/5/10 Ian Smith <smithi at nimnet.asn.au>
> [..]
>  > > cpu0: <ACPI CPU> on acpi0
>  > > est0: <Enhanced SpeedStep Frequency Control> on cpu0
>  > > p4tcc0: <CPU Frequency Thermal Control> on cpu0
>  > > cpu1: <ACPI CPU> on acpi0
>  > > est1: <Enhanced SpeedStep Frequency Control> on cpu1
>  > > p4tcc1: <CPU Frequency Thermal Control> on cpu1
>  > > =======
>  > >
>  > > Right, it's a Core2 Duo at 2GHz, using est (absolute) frequency
> control,
>  > > and p4tcc (relative) frequency thermal control.  See cpufreq(4) ie
> 'man
>  > > cpufreq'.  This combination seems to be a problem for these in some
>  > > recent machines.  While at 250MHz it should be using only 4.375 watts,
>  > > that doesn't seem to be working right on these, and you may be better
>  > > off (regarding heat) using the fewer frequencies provided just by est.
>  > >
>  > > =======
>  > > dev.cpu.0.freq: 250
>  > > dev.cpu.0.freq_levels: 2000/35000 1750/30625 1500/26250 1250/21875
>  > > 1000/17500 750/13125 500/8750 250/4375
>  > > dev.cpu.0.cx_supported: C1/1 C2/57
>  > > dev.cpu.0.cx_lowest: C1
>  > > dev.cpu.0.cx_usage: 100.00% 0.00% last 500us
>  > > =======
>  > >
>  > > So, cpufreq seems only to be seeing your 2000 frequency, and each of
>  > > these other frequencies are 7/8, 6/8 .. 1/8 of 2000MHz, generated by
>  > > p4tcc, which seems odd, unless 1000 is a 'real' supported frequency?
>  > >
>  > > Please try adding the following lines to your /boot/loader.conf and
>  > > rebooting, then paste what 'sysctl dev.cpu' says after that?  Also
> check
>  > > if just doing this makes any difference to your overheating issue, as
> it
>  > > has for some people.
>  > >
>  > > hint.p4tcc.0.disable="1"
>  > > hint.acpi_throttle.0.disable="1"
>  > > hint.p4tcc.1.disable="1"
>  > > hint.acpi_throttle.1.disable="1"
>  > >
>  > > You need both pairs; if you just disable p4tcc, acpi_throttle would
>  > > attach and provide those same frequencies, so would be no advantage.
>  > >
>  > > You may also find 'Fighting for the power' by Alexander Motin useful,
> in
>  > > particular enabling your available C2 state to reduce idle power use:
>  > >
>  > >
> http://lists.freebsd.org/pipermail/freebsd-current/2009-May/006436.html
> [..]
>
>  > Ok, thanks for it. The topic of Alexandre is very useful! So, i modify
> my
>  > loader.conf an rc.conf file by adding lines you told me to write
>  > (loader.conf) and lines to enable C2 and others to reduce number of
> sounds
>  > genertaed interrupts (see 'Fighting for the power' of Alexandre).
>
> Yes C2 should help overall, even on CPU-intensive tasks like buildworld,
> and advice on how to to power-down subsystems you're not using is handy.
>
> Is there any mention in your BIOS settings about 'C1E' or similar?
>
>  > I reboot and as you have asked me, the output of systcl dev.cpu after
>  > reboot : http://pastebin.com/bajvzy1W
>
>  : dev.cpu.0.freq: 250
>  : dev.cpu.0.freq_levels: 2000/35000 1750/30625 1500/26250 1250/21875
>  :  1000/17500 750/13125 500/8750 250/4375
>  : dev.cpu.0.cx_supported: C1/1 C2/57
>  : dev.cpu.0.cx_lowest: C2
>  : dev.cpu.0.cx_usage: 5.51% 94.48% last 229us
>
> Well now I'm confused, unless your C2D actually is really providing all
> of those frequencies out of the box, without any throttling driver?
>
> Are you sure that after booting with those hints in /boot/loader.conf
> (check: in /boot directory) that no lines at all are shown matching:
>
>  % egrep 'p4tcc|acpi_throttle' /var/run/dmesg.boot
>
> If that shows nothing, then I've been way off-base about this ..
>
>  > I tried to see how to behave my laptop (regarding fan noise and heat).
>  > Overall, i think it's a bit better, less noisy but not really perfect
> (sorry
>  > about my vagueness).
>  >
>  > I didn't do all steps of the topic of Alexandre and maybe after do them,
> it
>  > will be better again...
>
> First, a little (understandable) confusion; the 'Fighting for the power'
> post was by AlexandER Motin (mav at freebsd.org), the developer of recent
> snd_hda and mods to the powerd algorithms, among other things.  It was
> AlexandRE Kovalenko, note spelling, (cc'd), who replied offering to look
> through your BIOS ASL code if you dump it out and post it somewhere, as
> detailed on the ACPI debugging page.
>
>  > I tried to do an high cpu task but same behavor, system shuting down
> (heat
>  > exceed 92°C, critical...)
>
> Hmm, there was no 92C mentioned before; _HOT was 90C and I don't think
> FreeBSD uses _HOT at all, just _PSV and _CRT (any corrections welcome)
>
>  > I will to learn more about acpi debug and ASL from the handbokk and
> share
>  > you more about these.
>
> Good idea.  Generating your ASL is simple using those instructions, and
> Alexandre has a track record at spotting problems, especially thermal,
> among others who might be interested in looking through it.
>
> But please confirm there's no mention of p4tcc or acpi_thermal in dmesg?
>
> cheers, Ian

Hey ! My laptop is sweating! :)

So, after reading different manuals and others threads on the interdet, i
tried to tunn a bit my loader.conf with sysctl values but there are
somethings i definitly not understand.

Fist of all, Ian, i do what you told me to do about p4tcc or acpi_thermal
mention in my dmesg.boot file. No mentions found.

Next, i noticed that my temp increases until 80C but not higher,
significative?

Now, about the _PSV and _CRT values. I can set these but i don't what value
i have to use for these... Someone can advise me on that, regarding my proc
spefications.

Then, i generated my asl file and tried to compile it (as explained in the
handbook) to detect some errors but no errors, only warnings and optizations
and i will post it on the freebsd-acpi mailing list or somewhere else where
someone might be abble to read it and detect something wrong.

However (and it s a bit funny ;)) : i've got an panic after around 4min :

dev = ad4s1f, block = 1, fs = /usr
> panic: ffs_blkfree: freeing free block
> cpuid=1
> Uptime 4m41s
> Cannot dump. Device not found or unavailable

 Further, i can't shutdown or reboot my laptop, it block on 'Uptime XX' and
it never done...

I see one sysct value on the acpi manual but i've not tried yet.

I enabled the acpi debug (only error) but i don't know where i can read this
output... (don't find this information in the manuals).

To conclude, i'm a bit confused and i don't what is the problem and where i
have to go deeper, acpi bug? Maybe unable acpi and see what happend can be
interessting? (wait your opinion before do it).

After all, this problem is a good way to learn more about FreeBSD arch and
linux overall. Train my english and develop the habit to use the
documentation (i.e man page and books).

When i will have the suitable materials, i will try to 'clean' my fans but i
don't think i have this kind of problem.

PS : i don't really know i can consult the BIOS options and i don't respond
you about C1E but i see something really wirred. In my BIOS system overview,
my model of laptop is dv5 notebook PC oO, it's not true! So, i wait your
feedbacks about that, it's maybe the cause of all my problems? I hope (or
not).