> Date: Mon, 06 Jul 2009 15:04:51 +0200 > From: Aragon Gouveia <aragon_at_phat.za.net> > Sender: owner-freebsd-current_at_freebsd.org > > Hi, > > Patrick Lamaiziere wrote: > >>> I would like an option to set the minimum allowed CPU speed, instead > >>> the use of the sysctl debug.cpufreq.lowest. > >> This is so that powerd doesn't take so long to spin up power in the > >> adaptive modes? > > > > If the speed is too low, my machine is not very interactive here > > (running KDE and all...). > > Have you tried disabling throttling (P4TCC)? Add these to your loader.conf: > > hint.p4tcc.0.disabled=1 > hint.acpi_throttle.0.disabled=1 > > When you reboot you'll notice a lot of frequencies missing from > dev.cpu.0.freq_levels. The remaining ones are those affected by EIST only. > > The difference? EIST controls CPU frequency *and* voltage. Throttling > controls only frequency, and is much less effective at saving power > (while being very effective at slowing your system down). If you > disable throttling and use just EIST, your system will be much more > responsive at a very small power cost. It's really even worse than this. Throttling/TCC don't actually change clock speed. Thy simply "skip" N of every 8 clock cycles. The result is that power and performance under load are reduced by exactly the same amount and the effect of throttling drops with CPU load and is zero on an idle system. EST and C3 (or higher when available) are a much bigger wins. I see little value in the use of TCC or throttling. > The power cost above can be made up by setting C-state to C2 or even C3. > > If you're running FreeBSD 8, another way of saving a lot of power on a > notebook is to power down USB devices. Most notebooks' webcams and > fingerprint readers are internally connected to the USB bus. Note that even having the USB drivers loaded runs up battery use and blocks C3 even if no devices are connected to the USB. This is fixed in current by the new USB stack, but in v7, I build the kernel on my laptop without USB and load as needed. > > Alexander Motin did a lot of testing a while ago. Take a look at this > thread: > > http://lists.freebsd.org/pipermail/freebsd-current/2009-May/006436.html -- 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 3751Received on Mon Jul 06 2009 - 21:36:18 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:51 UTC