On Mon, Dec 28, 2009 at 11:09:48PM -0000, Ted Harding wrote:

On 28-Dec-09 20:57:14, Chris G wrote:

...

While this isn't particularly Linux related it is PC related and I'd guess some people here may be knowledgeable in this area.

I'm trying to cut down my PCs' power consumption, one way has been to offload 'server' functions from my desktop machine to a much lower power server machine. That's now done and I'm looking at other ways to reduce power consumption.

I've been looking at the power consumption of my desktop machine and noticed that when in standby it consumes about 20 watts. So I checked how much it consumed when powered down completely (but still conencted) and it *still* consumed 20 watts. I.e. the power supply and motherboard even when completely idle take 20 watts. I suspect that this is mostly power supply losses.

I have met this sort of thing before (though a long time ago, and I don't remember all the details). It was something to do with the fact that switching on was done via "intelligence" on the motherboard, i.e. instead of having the power button closing the power circuit by mechanically closing a switch, pressing the power button simply makes a connection between two pins on the motherboard, which then sends a signal to the PSU which then turns itself on.

As a result, the motherboard (and associated PSU circuitry) are permanently taking power in order to remain in a states of "alertness" while waiting, tongues hanging out, for their Lord and Master to touch them with the Finger That Giveth Life.

If you want to suppress this, then of course switch off at the mains! Hoping this helps (it may be barking up the wrong tree, in your case).

Yes, I know that, however I think the power used at idle is mostly the 'overhead' in the power supply rather than the 'keeping awake' power of the motherboard. This is what I was trying to explain.

The fact that when the motherboard actually is 'running' (well, sort of) in standby mode the power consumed is near enough identical to when it's 'off' suggests to me that most of that 20 watts is losses before the motherboard. Hence I believe a different (but how to choose it) power supply will keep the motherboard in standby mode but won't use so much power.

On Tue, Dec 29, 2009 at 08:58:27PM +0000, Wayne Stallwood wrote:

Chris G wrote:

...

When powered up, even though it's quite a fast machine (quad core processor, 8Gb of memory, two big disk drives) it only takes 50 watts so the power supply isn't all that inefficient when it's working.

So how can I search for power supplies that don't take much power when completely idle? Typical 'efficient' 300 watt and 400 watt power supplies are better than 80% efficient but that means they may well still consume 20 watts when doing nothing. Efficiency *doesn't* necessarily equate to low 'off' power.

Yes, this is the issue with The "80Plus" certification, it doesn't take standby current into consideration and the lowest power state they even measure is 20% of rated load capacity.

... and given that most power supplies nowadays tend to be 400 watts or more but motherboards are actually getting more efficient so they run at 50/60/70 watts that 80Plus is a bit pointless.

I have seen various designs of SMPSU and a whole lot of them suffer from terrible inefficiencies like this, there are "nice" ways of doing it but all would add cost and have no marketable relevance until something like the "One Watt Initiative" sets it's sights on computer equipment manufacturers and OEM's

Even then there is an issue getting there, The soft power circuit on a mainboard draws very little on its own as it is really just pulling the green wire on the PSU ATX connector down to ground. However in addition to that you have cmos battery saver circuits (which provide the cmos battery voltage when the machine is plugged in but not powered up) and WoL (requiring the ethernet chipset to be powered up and holding a link) these will probably on their own require more than 1W of power before conversion.

Add to this mainboards that power usb devices even when the machine is "switched off", possibly so that they can make "wake on keyboard/mouse event" possible for USB HID's and you start to wonder why we left the 2 pole mechanical mains switch of the original AT design in the first place.

Of course the answer to that is convenience both in terms of PSU installation, features like WoL, and allowing the software to "almost" power down the machine completely. Although I think we should have stuck with the mechanical mains switch in addition to a soft power "standby" button (like on a TV set so it is clear that this isn't "power off" as many people currently think). I still think we need the soft power functionality both for operating systems that now like to install updates as they are shutting down and for IT departments that want to WoL machines in the middle of the night to do security updates or whatever. But I think that there should be a front panel provision for a proper "mains off" button as well.

All that said, one final point of consideration is that depending on your measurement equipment. You may be getting rather skewed results when measuring a SMPSU's consumption due to the slightly odd power factor profile of switch mode power supplies, this anomaly can be even more prevalent when a SMPSU is in standby mode, regardless of whether PFC is fitted or not.

Yes, I'm aware of that, I'm not entirely convinced about the 20 watts measurement but it's certainly more than one or two watts. I did try and confirm the power consumption by reading the current with a meter that claims to be 'true RMS' and that suggested (assuming 240 volts) that the standby consumption is around 16 watts. Of course measuring 'true RMS' current and assuming the voltage isn't perfect but it at least showed that my original power measurement wasn't miles out.