How to disable acpi thermal?

[Daniel Eischen]

On Thu, 21 Feb 2008, Alexandre "Sunny" Kovalenko wrote:

>
> On Wed, 2008-02-20 at 20:14 -0500, Daniel Eischen wrote:
>> On Wed, 20 Feb 2008, Alexandre \Sunny\ Kovalenko wrote:
>>
>>>
>>> On Wed, 2008-02-20 at 18:48 -0500, Daniel Eischen wrote:
>>>> On Wed, 20 Feb 2008, Alexandre \Sunny\ Kovalenko wrote:
>>>>
>>>>> I assume (possibly incorrectly) that 1) your CPU is capable of the
>>>>> frequency throttling and 2) you are using frequency governor of some
>>>>> sort (see cpufreq(4) for detail). If this is not the case, the change
>>>>> will not help.
>>>>
>>>> I don't know about 1):
>>>>
>>>>    CPU: Intel Pentium III (933.08-MHz 686-class CPU)
>>>>    Origin = "GenuineIntel"  Id = 0x686  Stepping = 6
>>>>    Features=0x383fbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,MMX,FXSR,SSE>
>>>>
>>>> and 2), no, I'm not using a frequency governor from what I can
>>>> tell.
>>>>
>>>>    $ sysctl -a | grep dev.cpu
>>>>    dev.cpu.0.%desc: ACPI CPU
>>>>    dev.cpu.0.%driver: cpu
>>>>    dev.cpu.0.%location: handle=\_PR_.CPU0
>>>>    dev.cpu.0.%pnpinfo: _HID=none _UID=0
>>>>    dev.cpu.0.%parent: acpi0
>>>>    dev.cpu.0.cx_supported: C1/0
>>>>    dev.cpu.0.cx_lowest: C1
>>>>    dev.cpu.0.cx_usage: 100.00%
>>>>
>>>>    $ sudo kldload /boot/kernel/cpufreq.ko
>>>>
>>>>    $ sysctl -a | grep dev.cpu
>>>>    dev.cpu.0.%desc: ACPI CPU
>>>>    dev.cpu.0.%driver: cpu
>>>>    dev.cpu.0.%location: handle=\_PR_.CPU0
>>>>    dev.cpu.0.%pnpinfo: _HID=none _UID=0
>>>>    dev.cpu.0.%parent: acpi0
>>>>    dev.cpu.0.cx_supported: C1/0
>>>>    dev.cpu.0.cx_lowest: C1
>>>>    dev.cpu.0.cx_usage: 100.00%
>>>>
>>>>
>>>>> Also, since I have sent you that change, I have learned that setting
>>>>> hw.acpi.thermal.user_override=1 and hw.acpi.thermal.tz0._PSV=85C might
>>>>> accomplish the same thing as the ASL change. I saw it working for the
>>>>> thermal zone which already had sensible _PSV, but I have no hardware to
>>>>> try this approach when _PSV is not present in the ASL.
>>>>
>>>> Well, this is a server board, not a laptop, so I'm not sure
>>>> it even has CPU throttling.
>>>>
>>> As far as I can tell from the stuff above -- it does not. I was confused
>>> by presence of some pieces in the ASL, which would normally indicate
>>> that it does. One thing, which surprises me, is that with the lack of
>>> throttling and, what appears to be single speed fan or even a heatsink,
>>> I do not see how its thermal mode could behave differently in the
>>> presence of the load. Could you, please, send me output of the sysctl
>>> machdep.
>>
>> $ sysctl machdep
>> machdep.enable_panic_key: 0
>> machdep.adjkerntz: 0
>> machdep.wall_cmos_clock: 0
>> machdep.disable_rtc_set: 0
>> machdep.conspeed: 9600
>> machdep.gdbspeed: 9600
>> machdep.conrclk: 1843200
>> machdep.disable_mtrrs: 0
>> machdep.guessed_bootdev: 2687500288
>> machdep.cpu_idle_hlt: 1
>> machdep.prot_fault_translation: 0
>> machdep.panic_on_nmi: 1
>> machdep.kdb_on_nmi: 1
>> machdep.tsc_freq: 933073074
>> machdep.i8254_freq: 1193182
>> machdep.acpi_timer_freq: 3579545
>> machdep.acpi_root: 1009936
>>
> OK, I give up. The only positive idea so far is that your CPU needed
> these idle moments (machdep.cpu_idle_hlt: 1) to keep temperature under
> the critical level. AFAICR 4.x did not know much of anything about ACPI,
> so it did not exhibit any problems.
>
> To check whether this is the hardware issue or that of FreeBSD, you can
> boot from some recent Linux LiveCD and run 'openssl speed aes -multi
> 256', or some other CPU-intensive thing of you choice, and
> 'cat /proc/acpi/thermal_zone/TZC0/temperature' periodically to see
> whether temperature would raise as rapidly. It might not be a bad idea
> to do the same (former) thing under FreeBSD to make sure that it does
> indeed cause thermal shutdown ;)

I've done this, to somewhat ambiguous results.  I used OpenSUSE
Live CD (V10.3 / KDE).  So it booted into KDE just fine, and
'cat /proc/acpi/thermal_zone/TZC0/temperature' showed the temp
ranging from 98C to 100C.  When I ran 'openssl speed aes -multi 256'
it (Linux) showed the temperature climb to as high as 104C,
but still seemed to hover around 101-103C.  It didn't shoot
up much from the 99C.  But all that was with KDE, so I don't
know what the temp was without KDE running.

In FreeBSD, using 'openssl speed aes -multi 256', the temp
can climb as high as 117C (which is over the 110C _CRT
cap).  Since this is a server, I don't have any window
manager running on it.

> Another (and more dangerous) experiment would be to set
> hw.acpi.thermal.user_override=1 and hw.acpi.thermal.tz0._CRT=150C and
> see whether temperature would stabilize at some point above the old _CRT
> value. This might cost you a CPU, though...

I've already done that.  It stablizes at about 117 or so.
Since there's not much more we can do, I'll just drop the
issue.  Maybe I'll see if I can get new server to replace
it, kind of a shame since it's got much more horsepower
than it currently needs.

-- 
DE