[RFC] Patch to enable temperature ceiling in powerd

Johannes Dieterich dieterich.joh at googlemail.com
Wed Feb 20 21:51:43 UTC 2008


First, thanks for your reply!

Kevin Oberman wrote:
>> Date: Wed, 20 Feb 2008 22:10:41 +0100
>> From: Johannes Dieterich <dieterich.joh at googlemail.com>
>> Sender: owner-freebsd-acpi at freebsd.org
>>
>> Alexandre "Sunny" Kovalenko wrote:
>>> On Sun, 2008-02-17 at 14:40 +0100, Johannes Dieterich wrote:
>>>> Hi,
>>>> some not so nice news:
>>>>
>>>> That still holds true. Unfortunately portupgrade gcc overheats it again.
>>> You might want to do
>>>
>>> sysctl hw.acpi.thermal.user_override=1
>>> sysctl hw.acpi.thermal.tz1._PSV=85C
>>>
>>> and see if this gets you through the gcc compilation.
>> for a long long time it looked very good. But then it again overheated.I
>> might want to stress again that it happened out of a sudden. sysctl
>> dev.cpu reports 83-87 degrees for a long time and then it SUDDENLY shuts
>> down saying it is over 127 degrees. The workload between those two data
>> points has not changed.
> 
> This is sounding like a hardware problem.
> 
> On almost all modern systems, the CPU temperature is read from a single
> junction on the silicon of the CPU. If it makes sudden, inexplicable
> jumps, this implies that either the junction on the chip or the support
> hardware on the mobo (this is analog stuff) is misbehaving.
> 
> It is possible that something is causing BIOS to handle the values
> incorrectly, but that would seem very unlikely to me.
In general, I am willing to believe these things. There is a but though.
The problem appeared when upgrading from 6.2 to 7.0. OK, hardware breaks
and coincidences happen. But I tested later with an openSUSE and a
Knoppix Live CD (OK, JUST a LiveCD) and a stress test and it never went
over 79 degrees. But it reproducibly happens with FreeBSD 7.0 and also
there just under load (which puts the temperature anyway to 85 degrees).
I do lack an install of a Linux to test these issues there because the
notebook is my productive system at the moment.

So do you think this issue would not show with a Linux LiveCD but an
install?

Is it worth in your opinion putting the hard drive out and finding
another one to install a quick SuSE/RedHat/whatever for testing?

Thanks,

Johannes


More information about the freebsd-acpi mailing list