Spurious thermal shutdowns on Dell Studio 1557

[Kevin Oberman]

> Date: Fri, 09 Apr 2010 15:13:45 +0100
> From: Bartosz Fabianowski <freebsd at chillt.de>
> Sender: owner-freebsd-acpi at freebsd.org
> 
> > Great.  I'd still like to know what your remaining frequencies are, as I
> > couldn't correlate your previous full set with the throttling N * 12.5%?
> 
> The remaining frequencies are:
> 
> 1597/35000 1463/31000 1330/27000 1197/23000 1064/19000 931/15000
> 
> Additionally, I do not remember how many CX states I had before - but I 
> am pretty sure that it was more than just one. Now, the only CX state 
> remaining is C1/3.
> 
> Interestingly, disabling either the p4tcc or the acpi_throttle driver on 
> its own does nothing to cut down on frequencies. Only when both are 
> disabled via hints does the list drop from 13 frequencies to the 6 above.

This is by design. TCC and throttling do the exact same thing, but TCC
does it better, so if only throttling is available , it is used and when
both are available, TCC is used.

Again, there techniques a=were designed for thermal management, not
power management. (TCC is Thermal Control Circuit) They are very close
to useless for power management and I have been campaigning for some
time to have them pulled from the FreeBSD power management system. EST
and similar tools actually do power management well.

Now, for a really weird suggestion, one that makes no sense, but works
perfectly on my laptop:
In /boot/loader.conf:
hint.p4tcc.0.disable="1"
hint.acpi_throttle.0.disable="1"
hw.pci.do_power_nodriver="2" (also, try 3, but my system hangs with 3)

Then build a kernel WITHOUT "device cpufreq" (or nodevice cpufreq)

I would expect this to provide NO frequency control, but I get (and use)
the five frequencies provided by EST: dev.cpu.0.freq_levels: 2000/27000
1600/22600 1333/19666 1066/16733 800/13800 

Testing has shown that this is a REAL power savings.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751