Slight OT: power consumption of PC PSUs

[Ed Sirett]

I love speculation/guesstimations 8-)

There is little change in the PF with load. I guess the PF is to a large extent an inherent property of the of the design of the PSU. Presumably the PF is _leading_ for SMPSs. The PF of the SPeakers is doubtless a X-former hence its poor PF.

You have only got to see the amount of fins and fan(ette)s in a newer sys box to realise where the energy is going.

I guess that although the circuits are physically smaller there are many more of them (does this work out as proprtional to the area of the silicon in the sys box?). Not that much more? Perhaps the difference is that the circuits are also switching much faster in the newer box.

[The Natural Philosopher]

Its that basically. CMOS certainly had very liitle current draw in ON or OFF states - not sure what technology is used - but anyway, the current all goes to charge and discharge the inherent micro capacitors in the chips. And it has to go through the transistors to do it, so they get effin hot.

WE never put heastinks onm 2Mhz Z80's mate.

>

[Dave Liquorice]

Far to many variables. Soem kit in "standby" takes just as much power when "on". Much easier to read the meter when you go to bed and when you get up noting the times as well do this for a week. Then use some simple maths to work out how much power you have used each night and an average.

[Andy Wade]

Yes, although the PF will, in general, tend to get worse for light loads.

The non-unity PF for SMPSs is mostly due to harmonic currents, rather than phase shift. Another way of looking at it is that the spiky current waveform that you get with a low impedance source (the mains) feeding a rectifier and capacitor input filter means that the RMS input current is higher than you'd need to feed the same average power to a resistive load, hence the VA is higher than the watts - i.e. the PF is less than unity.

Of course it's not quite so simple in practice nowadays because the EMC regulations effectively require SMPSs over 50W to have power factor correction.

Yes - it's a low-cost wall-wart designed only for low initial cost, with no consideration whatever of efficiency or running cost. I was contemplating throwing it away and extracting the 12V supply for the speakers from the system unit.

Other things being equal, the power is proportional to the clock frequency and to the square of the supply voltage -- hence the drive to lower and lower chip supply voltages.

[Andrew Gabriel]

I use Solaris x86 (not surprisingly), and with journalling filesystem, I've never had any problems with power cuts and no UPS. Funny enough, there seem to have been 4 power cuts yesterday. Prior to that, the systems had been up for around 170 days since previous power cut, which IIRC, was also last of a several power cuts over an hour or so.

I think you don't normally bother powering the systems off. You would program the UPS to shut the system down when UPS has less than, say,

2 minutes left, and the UPS itself shuts down some 2 minutes later. When power comes back, everything comes back on as it was switched on when power went off (ATX systems should remember this, although since none of my systems are newer than P120's things may have changed).

However, UPS software is notoriously crap, and it's not at all uncommon for it to be responsible for more downtime than power cuts are. When I setup some commercial servers, I eventually stopped using it (except for monitoring the UPS) -- system availability was much improved just by letting the UPS kill the power to the running systems when battery went flat. In your scenario, if the mains power comes back on after UPS has told systems to shutdown but before its own battery went flat, what causes systems to reboot?

[Andrew Gabriel]

Negligable leading factor, but mostly it isn't due to phase shift.

I would imagine a transformer with little/no secondary load, which looks like an inductor across the mains.

Yes, and there are so many more of them because in addition to the silicon getting bigger, the individual circuits get smaller.

[Witchy]

maths! eep :)

witchy/binarydinosaurs

[Ed Sirett]

Come to think of it. the 4W is probably almost entirly the iron and copper losses of the wall-wart. There's nearly 4 quid a year to be saved by powering the speakers from elsewhere.

[Dave Liquorice]

CPC have an ISA bracketed(*) card that will provide a regulated supply at a range of voltages and both polarities for less than a tenner. It gets its power from a drive power connector. Fitted to a couple of the machines here, speakers go on/off with the PC and there is one less wall wart.

(*) It doesn't connect to the slot edge connector so will fit a PCI based machine but the card is mounted on the other edge of the bracket compared to PCI so one has to watch where things meet or get close.

[David]

Are you saying that an ordinary wall-wart costs that much to run?

David (counting up the number of the damned things he currently has permanently plugged in around the house...)

[Dave Liquorice]

My ATX boxen don't, the good old 486 and P133 do. On the 486/P133 the power switch is a real power switch in mains lead when power returns up comes the PC. The perishing ATX boxes have a "soft" power switch I have yet to find away of getting them to power back up on a return of mains. There is a BIOS setting but it doesn't appear to have any effect, I suspect something has to be set during the shutdown sequence. When running doze and you shutdown the box switches "off" with Linux or OS/2 it doesn't.

The OS/2 stuff from APC isn't particulary good or wasn't 4 years ago. A friend wrote a far better bit of software which I use.

ISTR that there is a bit of dialogue between the UPS and the PC software. The UPS tells the PC of its status, on mains, on battery, battery low etc and the PC software acts on that information. If a shutdown is started it tells the UPS to switch off after a delay. The UPS will then go through it's programmed startup routine, I think, I've never really played with it as I'm normally here and it's no great shakes if stuff doesn't come back up automagically anyway but it would be nice.

[Andrew Gabriel]

ATX systems are supposed to save the soft power state to NVRAM and if mains goes off/on again, return to same state. So if system was running before power cut, it should reboot, whereas if system was soft powered off, it should stay that way (you may notice it come on for a split second whilst it reads the NVRAM). Some systems may have BIOS settings or ACPI functions to override this behaviour.

The ability for the OS to soft power down a PC requires it to be able to understand the BIOS ACPI function, developed mainly by Microsoft with the apparent intention that no one else would be able to make head nor tail of it, which is why it often doesn't work when using other operating systems. Prior to that, motherboards supported a sequence of some bytes which when written to an i/o address in the right order and relatively quickly, would trigger the soft powerdown, although the sequences varied between motherboards.

APC for Unix had various faults -- leaked memory until the system ran out and hung (HP-UX), after a few weeks it would get stuck in a 100% CPU loop arguing with the serial port driver (Solaris and HP-UX), and it was responsible for a number of security alerts.

[Andy Wade]

Most of them aren't as bad as the one in question: 1.5 to 2 watts is more typical. 2 W times 8760 hrs/yr is 17.5 khW/yr, which would cost you about £1.20 per year at 7p per kWh.

[stefek.zaba]

Do the arithmetic. 5W means 200hours to consume one Unit (kWh). To a back-of-the-envelope approximation, that's a tad over a week (200 =

8*25, but only 24h and 7 days in a week), so let's say there's 50 such periods in a year (round number less than 52 rounding in the right direction to compensate for the overguesstimate previously). So the wallwart will eat about 50 Units in a year; at 7p a unit that's 3.50, at 8p it's the 4 quid mentioned. As others have said, 5W is more than most wall-warts will use when doing nothing, but poorly-designed or higher-output ones could pull this much. (One comprehensive report with per-appliance breakdowns and considerations for overall generation and equipment-production policy comes from the Land Down Under, at summary figure there is "Average standby and miscellaneous consumption was 86.8 Watts... in Australia this equates to 11.6% of [year] 2000 residential electricity use".)

Powering *off* your a-v kit, multiple little peripherals, and the like is a Win in electricity bills and overall energy consumption (the latter is muddied in a temperate/cold climate as effectively all the "wasted" power goes to warm up the house, decreasing the amount of deliberate heating you have to do, most typically from a cheaper-to-you source

- gas CH - but one which contributes a midgeon more to carbon dioxide emissions than the balance of electricity generation sources which include nuclear and hydro-electric). For kit which spends a small part of its time being actively used it's a win in life expectancy too; for frequent-use items we're into the more complex tradeoffs about switching surges versus resting...

Stefek

[Ed Sirett]

A good rule of thumb is that 1W on 24/7 for a year is around 1 quid.

1Wyear is about 31MJ roughly 8 kWhr. i.e about 60p So I'm a bit out but in the right area.

-- Ed Sirett - Property maintainer and registered gas fitter. The FAQ for uk.diy is at

Gas fitting FAQ CH FAQ

[John Laird]

It's easier to work out 1W for 24hrs/day for 40 days being more or less exactly 1kWh. Multiply by 9 and you have 1W for 360 days equating to 9kWh.

60-70p sounds about right.

[David]

Hmm - just musing on this; when I set up my PC workstation I extended a ring main within a stud partition to provide 3 double sockets side-by-side(not very accessible, under the desk) for all the computer gear; thus I have wall-warts there for printer, broadband network router, speakers, emergency modem, phone charger... so it would be nice to be able to just switch off all 3 sockets via a wall switch at desktop height.

How could I do this? The 2 ring main cables come down from the ceiling within the same section of the partition. *Theoretically* I could reconnect the ring without the sockets there, and re-wire them as a fused spur; however that would contravene regs wouldn't it. Three fused spurs, with separate switches maybe? Yuck. Surely there's a proper, neater way?

TIA David

[John Rumm]

Take the sockets out of the ring and connect them as a single spur to a switched fused outlet which is in the ring. Having the fuse makes this OK since the cable of the spur is now protected from overload by the fuse.

[chris French]

In message , Andrew Gabriel writes

this is certainly the behaviour here with this ATX system.

[Langis]

My washer states :-

Consumption in kWh - 0.95 Running time - 1h58m

So only ~14p per wash? I figured it would be much higher than that.

-- S i g n a l @ l i n e o n e . n e t