[Beowulf] OT: acpi-cpufreq won't load because???

[mathog]

This is off topic, but I suspect somebody here has been through this, so 
please bear with me.

Two Dell machines in the same room, both idle:

           Precision T7600      Poweredge T620
CPU       E5-2630 0 @ 2.30GHz  E5-2695 v2 @ 2.40GHz
kernels:  2.6.32-642.3.1       2.6.32-642.4.2
OS:       Centos 6.8           Centos 6.8
cpuMHz:   1200                 2400     (from /proc/cpuinfo)
cpuT:     40-50C               20-29C   (from sensors)

The first one seems to be running rather hot (assuming the temperature 
values lm_sensors lists are correct), especially given that it is at 
half the MHz, so I wanted to try cpufreq to see if it made any 
difference.  Neither had it loaded initially:

On the Precision this works:

   modprobe acpi-freq

and all the frequency control bits and pieces appear.  Everything as one 
would expect.

on the Poweredge though:

modprobe acpi-cpufreq
FATAL: Error inserting acpi_cpufreq 
(/lib/modules/2.6.32-642.4.2.el6.x86_64/kernel/arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.ko): 
No such device

on the poweredge at boot the only relevant message in dmesg is:

p4-clockmod: Warning: EST-capable CPU detected. The acpi-cpufreq module 
offers voltage scaling in addition to frequency scaling. You should use 
that instead of p4-clockmod, if possible.

grub.conf does not say anything, one way or the other about acpi. I have 
not been able to get into the BIOS on the poweredge yet to see if 
something is turned off there.   Both machines show:

dmidecode | grep -i acpi
                 ACPI is supported
                 ACPI (ACPI supported)
                 ACPI (ACPI supported)

biosdecode shows:

poweredge:
ACPI 2.0 present.
         OEM Identifier: DELL
         RSD Table 32-bit Address: 0x000F0C94
         XSD Table 64-bit Address: 0x00000000000F0DB8

Precision:
ACPI 2.0 present.
         OEM Identifier: DELL
         RSD Table 32-bit Address: 0x0DFFEF18
         XSD Table 64-bit Address: 0x000000000DFFEE18

The same failure to load acpi-cpufreq happens on yet another machine, a 
Poweredge T310 with a Xeon X3470.  This one has CPU temps in the 20-30C 
range, like the other poweredge, unlike the Precision.

Any idea why the poweredge refuses to load acpi-cpufreq?  (Wait!  Found 
a manual for a poweredge and the BIOS apparently has a setting that in 
one state lets the BIOS control the CPU speeds, and in the other lets 
the OS do it.  I bet it must be in the first state on those two 
Poweredges.  What does that setting do with respect to fan speed control 
though?  Turn it to "OS" do the fans blow full speed all the time?  How 
does the Dell CPU control compare with cpufreq?)

For bonus points, any idea why the Precision cpus are so much hotter?

For extra extra bonus points - what magic incantation is needed to read 
the fan speeds with lm_sensors on a Dell?  It doesn't show up on either 
of these machines.  ipmitools on the poweredge will show 2 fan speeds.  
That is not an option on the Precision though, as it has no ipmi.

Thanks,

David Mathog
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech