long flight; need battery power!

Philip S. Schulz ph.schulz at gmx.de
Fri Aug 26 17:38:14 GMT 2005


I'm neither Frank nor Colin but I think I can answer this anyways ;-)

Jeremie Le Hen wrote:
> Hi Frank, Hi Colin,
> 
> 
>>I recently installed sysutils/est and sysutils/estctrl (Enhanced
>>Speedstep Technology, kernel module and daemon) and it brought me
>>about 45 minutes more battery power.
>>Perhaps it's worth a try, if you don't already use it.
> 
> 
> Could you tell us what it brings in regards to the FreeBSD cpufreq
> framework please ?
> 

  est(4) has been reworked and integrated into cpufreq(4). The 
functionality of estctrl(1) has been integrated into powerd(1). There is 
no benefit of using est(4)/estctrl(1) over cpufreq(4)/powerd(1). Ok, if 
your running anything older than 5.4-RELEASE, then there is no 
cpufreq(4) framework...

  There are, however, a few differences: est(4) is, in cpufreq(4) terms, 
an "absolute driver". cpufreq(4) however might detect other "relative 
drivers" like e.g. acpi_perf(4) and combine relative and absolute 
drivers to create the list of available frequencies. This has the 
effect, as Kevin Obermann has noted, that some apparently "lower" 
frequencies might actually use more power than a "higher" frequency. For 
instance, on my Pentium M 1.5 GHz, cpufreq(4) allows me to set the 
frequency to e.g. 800 MHz or 750 MHz. 800 MHz would be a native 
frequency of the CPU while 750 is .5 * 1500 MHz resulting from the 
combination with (I think) p4tcc(4).

  Personally, I find this very annoying b/c 750 MHz makes the fan spin 
up more often than 800 MHz where I don't feel any difference in performance.

  In order to work only with absulute drivers, i.e. emulate the est(4) 
behaviour, use /boot/device.hints to disable all relative drivers your 
system detects.

HTH,

Phil.

-- 
Don't fix it if it ain't broke.


More information about the freebsd-mobile mailing list