[Intel-gfx] OT: Power metering [Was: Re: [PATCH] add LVDS downclocking support]

[Steven J Newbury]

On Mon, 2009-03-23 at 11:05 +0000, Peter Clifton wrote:
> On Mon, 2009-03-23 at 00:30 +0000, Steven J Newbury wrote:
> > > [1]
> > > http://www.alamaison.com.au/p/361923/voltcraft-plus-energy-monitor-3000.html
> > I'm not surprised you didn't see any change measuring power usage at the
> > outlet.  Laptop power supplies are rarely switched mode, and typically
> > provide their full rating while in use.  They aren't typically designed
> > with energy efficiency in mind.
> 
> Actually, this isn't correct. Every laptop PSU is almost certainly a
> Switched mode unit, as is every PC PSU since probably the AT days.
> 
> SMPS, as other linear types will provide as much energy as demanded of
> them (within sensible bounds), and care is taken as to the efficiency of
> the design since this determines the heat transfer requirements from the
> unit. On laptop supplies, there isn't the luxury of using forced air
> cooling.
> 
Of course you are absolutely correct; it occurred to me this must be the
case as I posted.  One of those embarrassing posts to posterity, oh
well, it can join all the other times I've said things I wish I could
take back! :)
> I'd hazard a guesstimate of the main PSU unit's efficiency at 80-90%,
> Part / no load efficiency will be worse. I don't know how efficient
> laptop supplies tend to be when connected to AC but not charging the
> laptop (laptop off / on standby), but I could probably find out if
> anyone was really interested. (Our lab has a very precise calorimetric
> power measuring system, which has been used to verify standby powers for
> things such as mobile phone chargers and set top boxes).
> 
I think this is the critical point though, laptop chargers/PSUs just
aren't typically as energy efficient as desktop units, and are
necessarily simpler giving only a single output voltage, whereas desktop
units are able to adjust to demand for the specific requirements of the
machine for each voltage rail.  Certainly my current laptop PSU runs
pretty warm all the time, while I've had desktop units run cool to the
touch with the fan idled when the system is drawing little power.

> Also remember that there are also several switching PSU units within the
> laptop, including the one which steps the battery voltage down to the
> required levels. Those are probably designed with efficiency in mind,
> due to the thermal issues, and to increase battery life.
Yes absolutely.

> What is true though, is that you're looking for a fraction of a watt
> power change in a device pulling about 20 watts, via a mains adaptor
> which will draw harmonic currents well beyond 50Hz. The plug-top power
> meters are designed to measure up to about 13A, (~3kW). The unit you
> mentioned claims <1% accuracy of voltage and current readings, meaning
> its current measurement might be accurate 130mA. I'm not going to get
> into the statistics of whether the voltage / current measurement
> inaccuracies will be correlated, but lets say at best 1% accuracy in
> power measurement. That is corresponds to 30W.
My laptop consumes around 11W when idle with the screen backlight set to
low brightnesss, at that level of current draw with a 90W PSU (as in my
case) I would imagine the supply side inefficiencies dominate within the
PSU over a few % savings from the demand side.
> 
> In reality, the harmonics drawn by the switching PSU could easily throw
> off such a power meter if it doesn't sample V and I at a high frequency,
> with the V and I channel measurements well synchronised.
> 
Yes, this is definitely a significant potential source of error.  Like
blinking in sync with a strobe light ;)
> > If you want to measure the energy efficiency of a laptop, your best bet
> > is powertop when running from battery.
> 
> Long story short... yes, I concur. These measurements, whilst of unknown
> absolute accuracy and calibration, are almost certainly made on smooth
> DC power rails inside the laptop, and with sensing components scaled to
> match the expected range of powers drawn by the laptop.
Not that there isn't an independent value in measuring the full system
power usage when on AC, but significant efficiency improvements within
the laptop  could well be lost in the noise at the full system level as
the PSU becomes less efficient when supplying lower levels of current.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20090324/f7fa98d8/attachment.sig>