power saving

power saving

[Tania Oka]

I would like to reduce power consumption on a custom MPC855T board.  I
notice in arch/ppc/kernel/idle.c idled() that the power saving code
seems to be skipped for processors other than 603/750 types.  Can anyone
tell me why that is?  Is it not a good idea on the 855 for some reason?
Is there anyone who has tried this on the 855 (or 850 or 860)?  Any
comments on the general subject welcome.

Thanks


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: power saving

[Dan Malek]

Tania Oka wrote:
>
> I would like to reduce power consumption on a custom MPC855T board.

This has been asked before, and you can read about it in mailing
list archives.

Basically, the MPC8xx processors have a dynamic power management
system that doesn't gain much from just doing something in the idle()
loop.  The majority of the power is consumed by peripherals.  If they
are not enabled, they automatically power down.  All of the MPC8xx
drivers "do the right thing" on open and close to power up/down
peripherals.

If you have some very low power standby requirement, you need to add
lots of power management software to do things like power down/up
any external peripherals on your board, put DRAM in autorefresh
mode, shut down the clocks and put the processor into a deep sleep.
You could look at some of the PowerMac power management code for
design ideas.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: power saving

[August Hoerandl]

On Monday 07 May 2001 09:49 pm, Dan Malek wrote:
> Tania Oka wrote:
> > I would like to reduce power consumption on a custom MPC855T board.
>
> This has been asked before, and you can read about it in mailing
> list archives.

checked this - found nothing very specific

> Basically, the MPC8xx processors have a dynamic power management
> system that doesn't gain much from just doing something in the idle()
> loop.  The majority of the power is consumed by peripherals.  If they
> are not enabled, they automatically power down.  All of the MPC8xx
> drivers "do the right thing" on open and close to power up/down
> peripherals.
>
> If you have some very low power standby requirement, you need to add
> lots of power management software to do things like power down/up
> any external peripherals on your board, put DRAM in autorefresh
> mode, shut down the clocks and put the processor into a deep sleep.

yes - thats what i would like to do ;-)

> You could look at some of the PowerMac power management code for
> design ideas.

could anyone be a little more specific to get me to the right files ?

Regards
Gustl

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

power saving..

[mjstumpf]

I have a relatively large (11 drive) server that operates 24/7.

It really doesn't need to, though.  Use is sporadic at best; the purpose
is mostly archiving.

So what I'm wondering, is do we have auto-spin-down, power save features
available & STABLE?  This thing sucks down a lot of power, giving me a
bunch of heat in return :-)

Re: power saving..

[Peter L. Ashford]

On Mon, 25 Aug 2003, mjstumpf wrote:

> I have a relatively large (11 drive) server that operates 24/7.
>
> It really doesn't need to, though.  Use is sporadic at best; the purpose
> is mostly archiving.
>
> So what I'm wondering, is do we have auto-spin-down, power save features
> available & STABLE?  This thing sucks down a lot of power, giving me a
> bunch of heat in return :-)

The problem isn't the spin-down, but the spin-up.  When an I/O occurs, the
delay for the disks to spin up would be likely to cause the RAID card (if
you use one) to generate a timeout.  If you don't use a RAID card, it
might work (and it might not).

These disks probably draw around 10W each at idle.  This is a 110W load in
your system.  All things considered, I'd be more likely to want to put the
CPU into sleep mode.  This would probably save 60-80W per CPU (figure 80%
efficiency on the voltage conversion), with some small additional power
savings from the NorthBridge and memory.

If the power savings (or heat output) is REALLY important, you can leave
the entire system off, except when you need it.  This may decrease the
long-term reliability, especially of the power supply.

				Peter Ashford

Re: power saving..

[Mike Dresser]

On Mon, 25 Aug 2003, mjstumpf wrote:

> I have a relatively large (11 drive) server that operates 24/7.
>
> It really doesn't need to, though.  Use is sporadic at best; the purpose
> is mostly archiving.
>
> So what I'm wondering, is do we have auto-spin-down, power save features
> available & STABLE?  This thing sucks down a lot of power, giving me a
> bunch of heat in return :-)

I'd be concerned about premature disk failure from cycling the drives up
and down, which could end up being more expensive than your power bill.

Mike

Re: power saving..

[Gordon Henderson]

On Mon, 25 Aug 2003, mjstumpf wrote:

> I have a relatively large (11 drive) server that operates 24/7.
>
> It really doesn't need to, though.  Use is sporadic at best; the purpose
> is mostly archiving.
>
> So what I'm wondering, is do we have auto-spin-down, power save features
> available & STABLE?  This thing sucks down a lot of power, giving me a
> bunch of heat in return :-)

Sorry for a late reply, but I've been on holiday...

I asked something like this a while back and got a positive reply. I've
been using "noflushd" on a small server/firewall for some time with good
results.  It's only a 2-disk RADI-1 system and what I've found is that one
drive is spun-down nearly all of the time, and the other is spun down most
of the time. It onl seems to spin up one disk when it need to read
something thats not in memory. Spin-up time is about 8 seconds (these are
old drives)  and it works very well and runs very cool and mostly quiet.

Gordon
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
On Mon, 25 Aug 2003, mjstumpf wrote:

> I have a relatively large (11 drive) server that operates 24/7.
>
> It really doesn't need to, though.  Use is sporadic at best; the purpose
> is mostly archiving.
>
> So what I'm wondering, is do we have auto-spin-down, power save features
> available & STABLE?  This thing sucks down a lot of power, giving me a
> bunch of heat in return :-)

Sorry for a late reply, but I've been on holiday...

I asked something like this a while back and got a positive reply. I've
been using "noflushd" on a small server/firewall for some time with good
results.  It's only a 2-disk RADI-1 system and what I've found is that one
drive is spun-down nearly all of the time, and the other is spun down most
of the time. It onl seems to spin up one disk when it need to read
something thats not in memory. Spin-up time is about 8 seconds (these are
old drives)  and it works very well and runs very cool and mostly quiet.

Gordon

Power Saving

[Stephen Clark]

Hello List,

I am trying to get throttling to work on the following processor with 
linux 2.6.17-1.2142_FC4
with no luck.

AMD Athlon(tm) XP 1700+

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

The /sys/devices/system/cpu/cpu0/cpufreq/* directory does not exist. I 
have modprobed
the following cpufreq* modules.

Module                  Size  Used by
cpufreq_stats           6101  0
cpufreq_powersave       2241  0
cpufreq_ondemand        7009  0
cpufreq_conservative     7265  0


What am I doing wrong?

Thanks Steve

-- 

"They that give up essential liberty to obtain temporary safety, 
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty 
decreases."  (Thomas Jefferson)

Re: Power Saving

[Dave Jones]

On Mon, Nov 19, 2007 at 07:41:31PM -0500, Stephen Clark wrote:
 > Hello List,
 > 
 > I am trying to get throttling to work on the following processor

I think by throttling, you actually mean changing frequency/voltage ?
(throttling is something else, where the CPU skips every n cycles,
 which doesn't actually save any power)

 > with linux 2.6.17-1.2142_FC4 with no luck.

wow. that's a prehistoric kernel.

 > AMD Athlon(tm) XP 1700+

you lose.  Only the mobile athlons supported scaling their speed.
And even then, only if the BIOS supported it with the correct tables.
(Typically this means, "only laptops").

	Dave

-- 
http://www.codemonkey.org.uk

Re: Power Saving

[Stephen Clark]

Dave Jones wrote:

>On Mon, Nov 19, 2007 at 07:41:31PM -0500, Stephen Clark wrote:
> > Hello List,
> > 
> > I am trying to get throttling to work on the following processor
>
>I think by throttling, you actually mean changing frequency/voltage ?
>(throttling is something else, where the CPU skips every n cycles,
> which doesn't actually save any power)
>
> > with linux 2.6.17-1.2142_FC4 with no luck.
>
>wow. that's a prehistoric kernel.
>
> > AMD Athlon(tm) XP 1700+
>
>you lose.  Only the mobile athlons supported scaling their speed.
>And even then, only if the BIOS supported it with the correct tables.
>(Typically this means, "only laptops").
>
>	Dave
>
>  
>
well what about the info from /proc/

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

and:
cat /proc/acpi/processor/CPU0/throttling
state count:             2
active state:            T0
states:
   *T0:                  00%
    T1:                  50%

and:
[root@joker ~]# cat /proc/acpi/processor/CPU0/power
active state:            C2
max_cstate:              C8
bus master activity:     d18324c9
states:
    C1:                  type[C1] promotion[C2] demotion[--] 
latency[000] usage[01340140]
   *C2:                  type[C2] promotion[--] demotion[C1] 
latency[090] usage[02980043]

????
Steve

-- 

"They that give up essential liberty to obtain temporary safety, 
deserve neither liberty nor safety."  (Ben Franklin)

"The course of history shows that as a government grows, liberty 
decreases."  (Thomas Jefferson)

Re: Power Saving

[Dave Jones]

On Mon, Nov 19, 2007 at 08:05:29PM -0500, Stephen Clark wrote:
 
 > >I think by throttling, you actually mean changing frequency/voltage ?
 > >(throttling is something else, where the CPU skips every n cycles,
 > > which doesn't actually save any power)
 > >
 > well what about the info from /proc/
 > 
 > cat /proc/acpi/processor/CPU0/info
 > processor id:            0
 > acpi id:                 0
 > bus mastering control:   no
 > power management:        yes
 > throttling control:      yes
 > limit interface:         yes
 > 
 > and:
 > cat /proc/acpi/processor/CPU0/throttling
 > state count:             2
 > active state:            T0
 > states:
 >    *T0:                  00%
 >     T1:                  50%

See above. These are throttling states, used typically
if the system is overheating.

 > [root@joker ~]# cat /proc/acpi/processor/CPU0/power
 > active state:            C2
 > max_cstate:              C8
 > bus master activity:     d18324c9
 > states:
 >     C1:                  type[C1] promotion[C2] demotion[--] 
 > latency[000] usage[01340140]
 >    *C2:                  type[C2] promotion[--] demotion[C1] 
 > latency[090] usage[02980043]
 > 
 > ????

C states are unrelated to frequency/voltage scaling (which are P states)
They get taken advantage of automatically by ACPI as you can
see from the usage fields.

	Dave

-- 
http://www.codemonkey.org.uk

Re: Power Saving

[H. Peter Anvin]

Dave Jones wrote:
>  > 
>  > and:
>  > cat /proc/acpi/processor/CPU0/throttling
>  > state count:             2
>  > active state:            T0
>  > states:
>  >    *T0:                  00%
>  >     T1:                  50%
> 
> See above. These are throttling states, used typically
> if the system is overheating.
> 

And more importantly, there is no power advantages over T states as 
opposed to C states.  They pretty much mean pulsing the system in and 
out of either C1 or C2 depending on the CPU/chipset.

	-hpa

Re: Power Saving

[Phillip Susi]

H. Peter Anvin wrote:
> And more importantly, there is no power advantages over T states as 
> opposed to C states.  They pretty much mean pulsing the system in and 
> out of either C1 or C2 depending on the CPU/chipset.

Well, they FORCE the use of the C states to save power, even if the 
kernel doesn't want to because it still has runnable tasks.  This can be 
useful if you know you are running a task that busy waits but would not 
suffer from having the performance of the cpu cut in half.  Since it is 
busy waiting, the kernel never activates the C states, but the 
throttling will force it and thus, save you power and heat.