Invalid detection CPU frequency ?
lists at yazzy.org
Wed Sep 28 02:45:37 PDT 2005
On Wed, 28 Sep 2005 10:58:58 +0400
Vladimir Grebenschikov <vova at sw.ru> wrote:
> В вт, 27/09/2005 в 16:35 -0700, Nate Lawson пишет:
> > Vladimir Grebenschikov wrote:
> > > It looks strange but sometimes my notebook start on full-speed
> > > (1.7GHz) and sysctls output looks like:
> > >
> > > dev.cpu.0.freq: 1687
> > > dev.cpu.0.freq_levels: 1687/-1 1476/-1 1265/-1 1054/-1 843/-1
> > > 632/-1 421/-1 210/-1 dev.acpi_throttle.0.%desc: ACPI CPU
> > > Throttling
> > >
> > > And sometimes it starts on low speed (~500 Mhz)
> > > dev.cpu.0.freq: 595
> > > dev.cpu.0.freq_levels: 595/-1 520/-1 446/-1 371/-1 297/-1 223/-1
> > > 148/-1 74/-1 dev.acpi_throttle.0.%desc: ACPI CPU Throttling
> > > And speed can't be raised by powerd or other.
> > > Looks like it happens when notebook stats or shutdowns on
> > > batteries. Rebooting on AC power cures situation. But anyway - I
> > > do not think it is normal.
> > >
> > > Any suggestions ?
> > It's very simple. You have an old notebook that only supports
> > throttling, not SpeedStep or other options. Your BIOS sets the
> > initial speed based on AC line status on boot. If you look at your
> > CPU announcement at the top of dmesg, it will show 595 Mhz or
> > whatever since it uses a simple calibration loop.
> > I might consider always setting 100% in the attach routine for
> > acpi_throttle and p4tcc to get proper probe values. You'll still
> > have a problem if you use the TSC timecounter since it's only
> > calibrated at boot.
> Nate, thanks for explanation, is loading cpufreq.ko from loader is
> good solution ? Or I will still have TSC problems ?
I compiled it into my kernel which works fine on thinkpad R50e
device cpufreq #CPU frequency control framework
My entire setup can be found on:
More information about the freebsd-current