Save power at idle onTyan S2466N-MPX, athcool, etc.

Save power at idle onTyan S2466N-MPX, athcool, etc.

[David Mathog]

I'm currently on a quest to reduce the idle power
usage on various machines here.  For instance, using
"athcool" on a 2.6.11-12mdk kernel (Mandriva 2005LE)
on some Asus A7V266E machines it was possible to drop
the idle temps from 57.5C to 39C.

Flushed with success from that I next tried athcool
on a single processor (athlon-MP 2200+) Tyan 2466N-MPX
using a 2.6.8.1 kernel.  No joy.  It claimed to have
changed the bits it controls but this didn't do a thing
to the idle CPU temperature.  Neither did turning those bits
off, so it wasn't already in the low power state.

This gives a pretty good clue why:

%  modprobe processor
%  cat /proc/acpi/processor/CPU0/info
processor id:            0
acpi id:                 0
bus mastering control:   no
power management:        no
throttling control:      no
limit interface:         no

/proc/acpi/processor/CPU0/power
showed that only C1 was supported.  On the ASUS boards both C1 and
C2 were present.

So my questin is, has anybody managed to get the Athlon's 
"power saving at idle" mode working  on this board?  If so,
which BIOS settings, kernel, modules, and other magic was
needed to do so?  

Thanks,

David Mathog
mathog-7GExONQZ6ZKVc3sceRu5cw@public.gmane.org
Manager, Sequence Analysis Facility, Biology Division, Caltech


-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl

Save power at idle onTyan S2466N-MPX, athcool, etc.

[David Mathog]

I'm currently on a quest to reduce the idle power
usage on various machines here.  For instance, using
"athcool" on a 2.6.11-12mdk kernel (Mandriva 2005LE)
on some Asus A7V266E machines it was possible to drop
the idle temps from 57.5C to 39C.

Flushed with success from that I next tried athcool
on a single processor (athlon-MP 2200+) Tyan 2466N-MPX
using a 2.6.8.1 kernel.  No joy.  It claimed to have
changed the bits it controls but this didn't do a thing
to the idle CPU temperature.  Neither did turning those bits
off, so it wasn't already in the low power state.

This gives a pretty good clue why:

%  modprobe processor
%  cat /proc/acpi/processor/CPU0/info
processor id:            0
acpi id:                 0
bus mastering control:   no
power management:        no
throttling control:      no
limit interface:         no

/proc/acpi/processor/CPU0/power
showed that only C1 was supported.  On the ASUS boards both C1 and
C2 were present.

So my question is, has anybody managed to get the Athlon's 
"power saving at idle" mode working  on this board?  If so,
which BIOS settings, kernel, modules, and other magic was
needed to do so?  

Thanks,

David Mathog
mathog-7GExONQZ6ZKVc3sceRu5cw@public.gmane.org
Manager, Sequence Analysis Facility, Biology Division, Caltech



-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl

Re: Save power at idle onTyan S2466N-MPX, athcool, etc.

[Erik Slagter]

[-- Attachment #1: Type: text/plain, Size: 1192 bytes --]

On Fri, 2005-09-30 at 15:36 -0700, David Mathog wrote:

> Flushed with success from that I next tried athcool
> on a single processor (athlon-MP 2200+) Tyan 2466N-MPX
> using a 2.6.8.1 kernel.  No joy.  It claimed to have
> changed the bits it controls but this didn't do a thing
> to the idle CPU temperature.  Neither did turning those bits
> off, so it wasn't already in the low power state.
> 
> This gives a pretty good clue why:
> 
> %  modprobe processor
> %  cat /proc/acpi/processor/CPU0/info
> processor id:            0
> acpi id:                 0
> bus mastering control:   no
> power management:        no
> throttling control:      no
> limit interface:         no

We had a discussion about exactly this a few weeks ago.

It boils down to: the AMD768 chipset can do C1/C2/C3 but not out of the
box using ACPI. First because ACPI processor support did not support
multiprocessor, now it does, but the ACPI parameters of this motherboard
prevent C2/C3 from being used by ACPI processor support.

So, out of the box you have C1, using the amd76x_pm you also have C2 and
optionally C3 support (although that completely screws up my sata
controller).

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2115 bytes --]

Re: Save power at idle onTyan S2466N-MPX, athcool, etc.

[David Mathog]

> On Fri, 2005-09-30 at 15:36 -0700, David Mathog wrote:
> 
> > Flushed with success from that I next tried athcool
> > on a single processor (athlon-MP 2200+) Tyan 2466N-MPX
> > using a 2.6.8.1 kernel.  No joy. 

Erik Slagter replied:
> 
> We had a discussion about exactly this a few weeks ago.
> 
> It boils down to: the AMD768 chipset can do C1/C2/C3 but not out of the
> box using ACPI. First because ACPI processor support did not support
> multiprocessor, now it does, but the ACPI parameters of this motherboard
> prevent C2/C3 from being used by ACPI processor support.
> 
> So, out of the box you have C1, using the amd76x_pm you also have C2 and
> optionally C3 support (although that completely screws up my sata
> controller).

Got it, there are problems with ACPI and the S2466.  No big surprise
there since this board has all sorts of oddball problems.  (No WOL
etc.)

However, it sounded like you and others managed in the end
obtain some power savings at idle.  Correct?  If so,
could you or some other poster please explain how you accomplished
that, stating in particular:

1. relevant BIOS settings
2. minimum kernel required
3. any extra kernel patches required
4. any extra software (like athcool) required

The machines in question are single CPU beowulf nodes - they do
have an S3 video card but it is used only when there are
hardware problems and/or to change the BIOS.  So maybe
once every two years.  There's no sound, TV card, video streams,
or anything else like that.  These just sit around until a
compute job comes through, then they grind like crazy for a while.
It would be nice if they used less power while they were waiting.

Thanks,

David Mathog
mathog-7GExONQZ6ZKVc3sceRu5cw@public.gmane.org
Manager, Sequence Analysis Facility, Biology Division, Caltech


-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl

Re: Save power at idle onTyan S2466N-MPX, athcool, etc.

[Erik Slagter]

[-- Attachment #1: Type: text/plain, Size: 1882 bytes --]

On Wed, 2005-10-05 at 09:33 -0700, David Mathog wrote:

> > We had a discussion about exactly this a few weeks ago.
> > 
> > It boils down to: the AMD768 chipset can do C1/C2/C3 but not out of the
> > box using ACPI. First because ACPI processor support did not support
> > multiprocessor, now it does, but the ACPI parameters of this motherboard
> > prevent C2/C3 from being used by ACPI processor support.
> > 
> > So, out of the box you have C1, using the amd76x_pm you also have C2 and
> > optionally C3 support (although that completely screws up my sata
> > controller).
> 
> Got it, there are problems with ACPI and the S2466.  No big surprise
> there since this board has all sorts of oddball problems.  (No WOL
> etc.)
> 
> However, it sounded like you and others managed in the end
> obtain some power savings at idle.  Correct?  If so,
> could you or some other poster please explain how you accomplished
> that, stating in particular:
> 
> 1. relevant BIOS settings
> 2. minimum kernel required
> 3. any extra kernel patches required
> 4. any extra software (like athcool) required
> 
> The machines in question are single CPU beowulf nodes - they do
> have an S3 video card but it is used only when there are
> hardware problems and/or to change the BIOS.  So maybe
> once every two years.  There's no sound, TV card, video streams,
> or anything else like that.  These just sit around until a
> compute job comes through, then they grind like crazy for a while.
> It would be nice if they used less power while they were waiting.

Please search the mailing list archives for something like amd760,
amd768, amd76x_pm, smp, or simply my name. This has all been covered.

BTW I have had several stability issues with the board, so I am not
going to buy another Tyan board. My previous MSI board always worked
without any problem.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2115 bytes --]

Re: Save power at idle onTyan S2466N-MPX, athcool, etc.

[David Mathog]

Erik Slagter wrote:

> Please search the mailing list archives for something like amd760,
> amd768, amd76x_pm, smp, or simply my name. This has all been covered.

I already did that.  There is a lot of general discussion but
very few details on the final, working, configuration.  In fact,
having gone through that thread twice already I couldn't determine
for sure what the final, working configuration actually was!
That is what I'm after, in particular:

> > 1. relevant BIOS settings
> > 2. minimum kernel required
> > 3. any extra kernel patches required
> > 4. any extra software (like athcool) required

> BTW I have had several stability issues with the board, so I am not
> going to buy another Tyan board. My previous MSI board always worked
> without any problem.

Yeah, these boards are a bit of a pain.  I've got 20 of them.
Power them all down and turn them back on and the
odds are nil that they'll all come back up again.  One
or more always needs to be reset.  They can also get into a state
where reset is ignored, and since these are rack mounted,
the only way to regain control is to pull the power
supply cord out, wait a bit, and plug it back in again.

Moreover, two of them have failed so far.  But they didn't just
die outright, they started crashing for no good reason.  Perhaps
that's what you're seeing as "instability".  To its
credit Tyan did replace those boards under warranty.  The good ones
never crash, easily staying up for 6 months at a time.
(Under linux, with no graphics or other complex drivers
to confuse them.)  We use ECC memory, maybe that helps.

Regards,

David Mathog
mathog-7GExONQZ6ZKVc3sceRu5cw@public.gmane.org
Manager, Sequence Analysis Facility, Biology Division, Caltech


-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl

Re: Save power at idle onTyan S2466N-MPX, athcool, etc.

[Erik Slagter]

[-- Attachment #1: Type: text/plain, Size: 1630 bytes --]

On Thu, 2005-10-06 at 13:54 -0700, David Mathog wrote:

> > Please search the mailing list archives for something like amd760,
> > amd768, amd76x_pm, smp, or simply my name. This has all been covered.

> I already did that.

Okay, good.

> There is a lot of general discussion but
> very few details on the final, working, configuration.  In fact,
> having gone through that thread twice already I couldn't determine
> for sure what the final, working configuration actually was!

I my case: no configuration is working :-( That is because I use a
promise sata card which gets upset very easily.

I tried c3 but maybe if I limit to c2, it will work.

(all using amd76x_pm)

> That is what I'm after, in particular:
> 
> > > 1. relevant BIOS settings
> > > 2. minimum kernel required
> > > 3. any extra kernel patches required
> > > 4. any extra software (like athcool) required

1. none 
2. 2.6 iirc
3. amd76x_pm (http://www.sommrey.de/amd76x_pm/)
4. none

> Yeah, these boards are a bit of a pain.  I've got 20 of them.
> Power them all down and turn them back on and the
> odds are nil that they'll all come back up again.

Yep. They take ages to reset and then you'll have to have to luck it
actually does come up.

> (Under linux, with no graphics or other complex drivers
> to confuse them.)  We use ECC memory, maybe that helps.

Once running there is no problem (apart from the promise sata controller
issue, which very well may be a mainboard issue indeed). I have been
running with ECC for quite a while, but couldn't use it together with
the extra (non-ECC) I added later.

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2115 bytes --]