Re: High pitch noise when ACPI processor module is loaded.

Previous thread: Re: [PATCH 1/2] PM / Runtime: Add sysfs switch for disabling device run-time PM (rev. 2) by Alan Stern on Saturday, January 23, 2010 - 8:20 am. (2 messages)

Next thread: Re: [PATCH 23/24] drivers/block/floppy.c: Add function is_ready_state by James Kosin on Saturday, January 23, 2010 - 10:32 am. (4 messages)
From: Justin Piszcz
Date: Saturday, January 23, 2010 - 8:48 am

Hi,

Linux:
When I do not load the ACPI processor module, there is no high pitch 
noise.  When I load the ACPI processor module (so I can utilize Turbo 
Boost), Linux makes the system emit a high pitch noise.

Windows:
The same system, running Windows 7 does not emit any noise.

Is there something wrong with the ACPI processor module or its 
implementation on i5/i7 CPUs?

Justin.
--

From: Rafael J. Wysocki
Date: Saturday, January 23, 2010 - 1:12 pm

Messages about ACPI should be CCed to linux-acpi@vger.kernel.org (added).

--

From: Robert Hancock
Date: Saturday, January 23, 2010 - 1:42 pm

May not be anything "wrong", but some systems do have an issue with the 
voltage regulator circuitry where it can make noise with certain 
combinations of power draw and wakeup rates (causing rapid current draw 
fluctuations) where some of the components can resonate and make noise.

What does powertop report for this system? It might show some useful info.
--

From: Valdis.Kletnieks
Date: Sunday, January 24, 2010 - 10:34 pm

Sounds like singing capacitors.

http://lwn.net/Articles/145973/ - and see the last 2 comments.


From: Justin Piszcz
Date: Monday, January 25, 2010 - 2:26 am

Hi,

I think something else may be wrong?

I switched my kernel config to 100 HZ and the problem persists.

So the question remains, is it possible to utilize turbo boost (in Linux)
without the high pitch noise from the ACPI processor module?

Again, when the module is not loaded, no matter what HZ, there is no high 
pitch noise.

Justin.

--

From: Justin Piszcz
Date: Monday, January 25, 2010 - 2:56 am

Without ACPI processor module loaded, turbo boost never takes effect:

(turbostat)

  CPU   GHz    TSC    %c0    %c1    %c3    %c6   %pc3   %pc6   %pc7
    0   2.93   2.93   1.41  98.59   0.00   0.00   0.00   0.00   0.00
    1   2.93   2.93 100.00   0.00   0.00   0.00   0.00   0.00   0.00
    2   2.93   2.93   1.51  98.49   0.00   0.00   0.00   0.00   0.00
    3   2.93   2.93   0.97  99.03   0.00   0.00   0.00   0.00   0.00
    4   2.93   2.93   0.90  99.10   0.00   0.00   0.00   0.00   0.00
    5   2.93   2.93   0.34  99.66   0.00   0.00   0.00   0.00   0.00
    6   2.93   2.93   0.40  99.60   0.00   0.00   0.00   0.00   0.00
    7   2.93   2.93   3.07  96.93   0.00   0.00   0.00   0.00   0.00

With the module loaded, it works but there is a high pitch noise, again,
only under Linux, when the same benchmark (bzip2 a file) is run under windows
there is no high pitch noise:

  CPU   GHz    TSC    %c0    %c1    %c3    %c6   %pc3   %pc6   %pc7
    0   3.30   2.93   0.44   0.82  25.29  73.44   0.00   0.00   0.00
    1   3.33   2.93   0.24   1.02  25.29  73.44   0.00   0.00   0.00
    2   3.43   2.93   1.84   5.41  48.65  44.10   0.00   0.00   0.00
    3   3.41   2.93   0.49   6.76  48.65  44.10   0.00   0.00   0.00
    4   3.58   2.93  99.44   0.56   0.00   0.00   0.00   0.00   0.00
    5   3.43   2.93   0.03  99.97   0.00   0.00   0.00   0.00   0.00
    6   3.09   2.93   0.14   0.23  89.74   9.89   0.00   0.00   0.00
    7   2.13   2.93   0.01   0.36  89.74   9.89   0.00   0.00   0.00


Justin.


--

From: Ian Campbell
Date: Monday, January 25, 2010 - 3:10 am

I have to use the max_cstate=2 option to processor.ko in order to not
have a similar high pitch noise. I suppose you might get away higher
values of max_cstate depending on your hardware. Battery life doesn't
seem to be too adversely impacted.

Ian.

-- 
Ian Campbell
Current Noise: Ephel Duath - Ruins (Deep Blue And Violet)

Hmm...  Which would do a better job at driving physicists crazy?  Travel
faster than light, or a floating-point boolean value?
		-- Michael Mol

--

From: Justin Piszcz
Date: Monday, January 25, 2010 - 1:46 pm

Hello,

I tried this, but then the max speed bin is only 3.2GHz instead of 3.6GHz,
it does solve the problem but limits the CPU quite a bit, see below:

  CPU   GHz    TSC    %c0    %c1    %c3    %c6   %pc3   %pc6   %pc7
    0   3.20   2.93   0.33  99.67   0.00   0.00   0.00   0.00   0.00
    1   3.20   2.93   0.09  99.91   0.00   0.00   0.00   0.00   0.00
    2   3.20   2.93 100.00   0.00   0.00   0.00   0.00   0.00   0.00
    3   3.20   2.93   0.95  99.05   0.00   0.00   0.00   0.00   0.00
    4   3.20   2.93   0.08  99.92   0.00   0.00   0.00   0.00   0.00
    5   3.20   2.93   0.02  99.98   0.00   0.00   0.00   0.00   0.00
    6   3.20   2.93   0.06  99.94   0.00   0.00   0.00   0.00   0.00
    7   3.20   2.93   0.01  99.99   0.00   0.00   0.00   0.00   0.00

--

It would be nice if it were possible to find out what is causing it exactly
in the acpi cpufreq module, as this problem does not occur in Windows 7 and
I can achieve 3.6GHz in Windows 7 w/out any high pitch noise.

Justin.

--

From: Sitsofe Wheeler
Date: Monday, January 25, 2010 - 3:38 pm

That may well be related to the interrupt frequency the kernel is
running with. I remember reading that a lower HZ were less likely to
cause chip whine on some machines...

http://www.thinkwiki.org/wiki/Problem_with_high_pitch_noises#Change_the_timer_interrup...

-- 
Sitsofe | http://sucs.org/~sits/
--

From: Henrique de Moraes Holschuh
Date: Monday, January 25, 2010 - 5:33 am

Well, I am tempted to say you should consider getting a higher quality
motherboard (or, if it is just yours that sing, get it replaced on warranty
as that would mean you got one with bad components).

That said, if Windows can get that motherboard to shut up, so should Linux.
You just need to find the right knob to push that will make the power draw
behaviour more like whatever Windows is doing.

The first good bet is to limit core C-state depth, e.g, to C3 (instead of
C5/C6), and if thet doesn't work, to C2.  After that, you can try limiting
package C-states if your BIOS will let you (I don't know if we can do that
distinction in Linux right now... do we?), play with cpufreq limiting the
core frequency changes, etc.

And of course, try all scheduler frequencies, with and without NO_HZ.

What motherboard is this?

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh
--

From: Justin Piszcz
Date: Monday, January 25, 2010 - 5:41 am

Intel's DP55KG Extreme motherboard, yes I will try changing the c-state 
later and hopefully that will fix the problem, thanks.  Again, the board 
never makes any noise in the BIOS or Windows or Linux, only when you load 
the cpufreq-processor module.

Justin.
--

Previous thread: Re: [PATCH 1/2] PM / Runtime: Add sysfs switch for disabling device run-time PM (rev. 2) by Alan Stern on Saturday, January 23, 2010 - 8:20 am. (2 messages)

Next thread: Re: [PATCH 23/24] drivers/block/floppy.c: Add function is_ready_state by James Kosin on Saturday, January 23, 2010 - 10:32 am. (4 messages)