[NTLUG:Discuss] Temperature Triggers

Gilbert Morrow gkfmorrow at gmail.com
Sun Aug 30 11:17:45 CDT 2015


https://ask.fedoraproject.org/en/question/31514/how-to-change-cpu-temperature-thresholds-and-control-fan-speed/,
found this with Google search. Might help.
On Aug 30, 2015 10:17 AM, "Stephen Davidson" <gorky at freenet.carleton.ca>
wrote:

> Greetings.
>
> This is an older bios -- there are NO temperature settings, readings, or
> other temperature controls in the BIOS.
>
> Lm_sensors had some kind of "upgrade" at some point and now if a CPU
> temp (and/or GPU, depending on system capabilities) exceeds thresholds,
> it issues an "shutdown -h now" command to the OS.  For a 'modern' CPU,
> the thresholds are actually a bit on the high side.  For this particular
> line of CPUs, that threshold is low (for the parent or previous
> generation CPU, VERY low -- the threshold is just above it's normal IDLE
> temp of 85C!).
>
> There are a couple of processes that will run and load the CPU (one of
> which is the 'Up-to-date check'), and if they run long enough (or back
> to back), CPU will start approaching normal max temp (if memory serves,
> that's 99c on this system -- but it's been a while & HP's docs says
> 95c).  Air flow is good, and I've just disassembled and cleaned the
> system so all dust bunnies are removed -- and I did double check that
> the fans are running properly.
>
> Contents of 'sensors.d/laptop' file (note: when temp1_input ==
> temp1_crit, 'shutdown -h now' gets issued by a daemon for lm_sensors);
>
> chip "acpitz-virtual-0"
>    compute temp1 @+10, at +10
>
> Output:
> steve at sda64 ~ $ sensors -u
> acpitz-virtual-0 Adapter:
> Virtual device temp1:
>     temp1_input: 60.000
>     temp1_crit: 100.000
>
> k8temp-pci-00c3
> Adapter: PCI adapter
> Core0 Temp:
>     temp1_input: 44.000
>
> Without the sensors.d file;
> steve at sda64 ~ $ sensors -u
> acpitz-virtual-0 Adapter:
> Virtual device temp1:
>     temp1_input: 50.000
>     temp1_crit: 90.000
>
> k8temp-pci-00c3
> Adapter: PCI adapter
> Core0 Temp:
>     temp1_input: 44.000
>
> I need to be able to raise the temp1_crit to at least 95 (preferably 97)
> w/o also raising the temp1_input value.  Unfortunately, I found in my
> experiments that only the first @+10 is getting used in that compute
> line.  So, what did I miss in the documents for lm_sensors on this, or
> who do I need to contact to get this looked into?
>
> Regards,
> Steve
>
> On 08/27/2015 04:38 PM, Gilbert Morrow wrote:
> > BIOS controls the CPU temperature threshold, OS uses it to determine
> > readings.
> > I use to overclock a lot in the past, I used water cooling for the old
> game
> > boxes. I still have one for an old AMD processor, radiator, fan, and
> > heatsink water block.
> > On Aug 27, 2015 3:09 PM, "Stephen Davidson" <gorky at freenet.carleton.ca>
> > wrote:
> >
> >> Greetings.
> >>
> >> I have an oldstyle AMD CPU, one of the first generations after the
> >> legendary "Cook an egg on it". (I tried Googling for one of the videos
> >> from that era, but no luck in the first couple of pages, sorry).
> >>
> >> Normal max operating temp is 95c, but the system does a protective
> >> shutdown at 90c.  I've tried modifying this in lm_sensors
> >> /etc/sensors.d, but so far no luck.  The current documentation seems to
> >> be not current with the current code.  I have an entry:
> >>
> >> compute temp1 @+10, at +10
> >>
> >> but it seems that:
> >>  - only the first @+10 is read and is used both for setting critical and
> >> adjusting display
> >>  - temp1 is now displaying 10c higher than other sensors.
> >>
> >> NOTE: There is generally a 5c temp difference between 'Core0' and Temp1,
> >> Temp1 being higher.
> >>
> >> Any suggestions?
> >>
> >> Regards,
> >> Steve
> >>
> >>
> >> _______________________________________________
> >> http://www.ntlug.org/mailman/listinfo/discuss
> >>
> > _______________________________________________
> > http://www.ntlug.org/mailman/listinfo/discuss
>
>
>
> _______________________________________________
> http://www.ntlug.org/mailman/listinfo/discuss
>


More information about the Discuss mailing list