CPU frequency doesn't drop below 1200MHz (like it used to)

[Ian Smith]

On Sat, 23 May 2015 17:40:26 +0300, Kimmo Paasiala wrote:
 > On Sat, May 23, 2015 at 5:15 PM, Ian Smith <smithi at nimnet.asn.au> wrote:
[..]
 > >  > It's an Intel Atom running amd64 version of FreeBSD stable/10:
 > >  >
 > >  > FreeBSD firewall.rdnzl.info 10.1-STABLE FreeBSD 10.1-STABLE #1
 > >  > r283292: Sat May 23 01:08:03 EEST 2015
 > >  > root at firewall.rdnzl.info:/usr/obj/usr/src/sys/GENERIC  amd64
 > >  >
 > >  > CPU: Intel(R) Atom(TM) CPU D510   @ 1.66GHz (1666.68-MHz K8-class CPU)
 > >  >   Origin="GenuineIntel"  Id=0x106ca  Family=0x6  Model=0x1c  Stepping=10
 > >  >   Features=0xbfebfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,DTS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE>
 > >  >   Features2=0x40e31d<SSE3,DTES64,MON,DS_CPL,TM2,SSSE3,CX16,xTPR,PDCM,MOVBE>
 > >  >   AMD Features=0x20100800<SYSCALL,NX,LM>
 > >  >   AMD Features2=0x1<LAHF>
 > >  >   TSC: P-state invariant, performance statistics
 > >  >
 > >  > Powerd was working on 10.1-RELEASE but stopped working after upgrade
 > >  > to 10-STABLE and nothing was changed in BIOS settings.
[..]
 > >  > However, reading the other replies to this thread I get the impression
 > >  > that powerd(8) doesn't actually save energy on this platform and I'm
 > >  > better off without it?
 > >
 > > No, I don't think that's correct; using deeper C-states is most likely a
 > > bigger win, but higher than needed CPU freq will still use extra power,
 > > so run hotter. `sysctl dev.cpu` will also reveal your C-state usage.
 > >
 > > Reason I'm pursuing this is that this change shouldn't hurt, but it will
 > > flush out those cases where people were only getting cpufreq due to use
 > > of a 'relative' cpufreq driver like p4tcc, unless EST's enabled in BIOS;
 > > I suspect yours may be one such case :)  If not, there's a bug to fix.

Seems _I've_ got a bug to fix; I need to stop assuming all modern Intel 
CPUs are going to make SpeedStep and/or deeper C-states available :(

 > Looking deeper into this it appears I don't have speedstep (EST)
 > support in the CPU it being a crappy Atom D510:
 > 
 > http://ark.intel.com/products/43098

Indeed.  It is rated at only 13W TDP, so relatively low power anyway.

 > This the full 'sysctl dev.cpu' output:
 > 
 > % sysctl dev.cpu

 > dev.cpu.3.cx_usage: 100.00% last 65712us
 > dev.cpu.3.cx_lowest: C1
 > dev.cpu.3.cx_supported: C1/1/0
[..]
 > dev.cpu.0.cx_usage: 100.00% last 3132us
 > dev.cpu.0.cx_lowest: C1
 > dev.cpu.0.cx_supported: C1/1/0
 > dev.cpu.0.%parent: acpi0
 > dev.cpu.0.%pnpinfo: _HID=none _UID=0
 > dev.cpu.0.%location: handle=\_PR_.P001
 > dev.cpu.0.%driver: cpu
 > dev.cpu.0.%desc: ACPI CPU
 > dev.cpu.%parent:

It doesn't even provide dev.cpu.0.freq, and has no deeper C-states 
('Idle States' on that page) available, so it looks like you may as well 
not bother running powerd.  Others maybe can offer better suggestions.
 
 > So I should keep those two hints in loader.conf to use p4tcc I guess?

If this is a desktop I'd just let it run flat out, ie disable p4tcc and 
acpi_throttle, have no cpufreq and forget powerd.

If it's a laptop and power consumption on battery matters to you, you 
could see if p4tcc's lower frequencies actually save any power much, by 
running 'powerd -v' in a terminal while testing with different loads, or 
if your 'acpiconf -i0' shows discharging rates in mA or mW, or both.

Sorry again for my poor assumption, and thanks for the data point!

cheers, Ian