Energy-hogging Linux, what to do?
Nicholas Bodley
nbodley at speakeasy.net
Sat Oct 22 14:53:48 EDT 2005
On Sat, 22 Oct 2005 12:23:22 -0400, Rich Braun <richb at pioneer.ci.net>
wrote:
> Have yet to figure out how much energy those econobox UPS units are
> eating up; power efficiency ratings on those are nowhere to be found.
I need to study further (APC "paper" on types of UPSes (more than you'd
expect)) to be sure, but I suspect that under normal conditions, the UPSes
of the sort we are likely to have are not drawing much power at all; they
are only trickle-charcing the battery, and that doesn't involve much
power. In our equipment, power ends up as heat, so if a device creates
significant heat, it's drawing significant power. (For a crude idea, reach
around to the exhaust fan(s) and feel the air. For one thing, it shouldn't
be particularly warm, but also, it requires power to create that heat!
Think of electric space heaters... As well, estimate the amount of air
being warmed.)
Thanks to some friends, I have a lovely curb-find (!) DEC-badged
Mitsubishi DiamondPro 21TX, and it does create a significant amount of
heat.
Devices with larger motors in them, such as washing machines, convert
much/most of the incoming power into mechanical energy, of course. Fan
motors in computers do, too, but they don't draw much of the total power.
Because our power quality is generally quite good by world standards, I'd
say it's safe to assume that voltage is about correct, so for rough
estimates, I think you'd only need to measure current, and multiply by
voltage, but, there's a "gotcha", there -- extreme distortion of the
current waveform in the power to a typical computer power supply that
doesn't have "power factor correction" to make it draw a sine wave of
current. That complicates estimating power. In Europe, afaik, these power
supplies are required to draw sine waves of current.
Typical power supplies have rectifiers directly connected to the line;
they have big capacitors, and no filter chokes. The charge in the caps. is
what actually runs the machine a good part of the time, and every
half-cycle, when the line voltage reaches its peak, the capacitors are
recharged -- "topped up", so to speak. This means that for a good part of
the AC cycle, only a very modest amount of power is being drawn, but at
the voltage peak, the current rises rather fast and drops off pretty fast
as a well.
Unfortunately, with AC, power isn't as simple as it is with DC; you have
non-sinusoidal (i.e., non-ideal) current (and maybe voltage) waveforms,
and sometimes current and voltage that are not in phase (such as with
electric motors)(that's poorer power factor). Of course, what counts on
your bill is what the utility meter reads; while consumer-owned portable
meters are not unheard-of, they are rare. In Waltham, at least, NStar
seem to be replacing old totally-electromechanical meters wholesale, so
used ones must be in good supply.
What I have in mind would be a properly-mounted meter on a stand or in a
box with a handle, a properly-wired substantial 3-wire power cord, and a
few properly-wired outlets.
Whether the pluggable electronic power totalizer ("Kill-A-Watt"?) is
accurate for typical computer power supply loads, I don't know; I'd guess
that it might be. It's not an especially new design, but might have been
updated.
I leave my machine on for long periods only when downloading .iso's;
otherwise, I power it down when done. Monitor stays in low-power standby.
I'm currently thinking of buying (cheaply!) a displayless laptop (user
removed the display, which went bad) to leave on all the time; laptops
have to draw less power. I'd turn off the monitor when not in use.
Btw, I know I'm risking being too technical; my apologies for that!
However, to explain better would triple the length of this message.
HTH, a bit.
Best regards,
--
Nicholas Bodley /*|*\ Waltham, Mass. (Not "MA")
The curious hermit -- autodidact and polymath
Retired electronic technician --
"test & troubleshoot to component level"
Happiness is a full Quabbin.
More information about the Discuss
mailing list