Ask Slashdot: Reliable Powering of ATX Systems?

David Muir Sharnoff writes in with this question: "ATX power supplies don't supply power unless the motherboard says to turn on. Many motherboards do not have a setting of: 'supply power all the time.' This means that ATX systems cannot safely be used a servers. Anyone know of an ATX power supply that ignores signal? " More appropriately: does anyone know of an ATX motherboard/power supply/case combination where powering is similar to AT systems?

WTX will appear *very* shortly.

WTX will replace ATX for server boxes. It appears to be a good spec to me.

WTX.ORG [wtx.org]

ATX specification URL

The ATX specification can be found here:

http://www.teleport.com/~atx/spec/index .htm [teleport.com]

(Boy I hope this message get moderated up.
That would be keen.)

ASUS P2BF & Intel Nightshade

I know for a fact that the ASUS P2BF has a "AC Power Fail Auto-Restart" BIOS option and the Intel Nightshade has a "Restore Power State" BIOS option that will turn the power on automatically if it was on when AC power was lost (i.e. power was not turned off via the front panel switch).

Mod to keep power on all of the time

Posted by el_steevo:

From the power supply wires coming from the power supply in the case to the MOBO, short the green wire to ground. This will give you power all of the time.

Use the force, luke... er... Lucas Digital?

This might be related to the Linux "use" at Lucas Digital that caused some discussion last week. Then again it might not and it just made me think of it.

If I had a rack of Linux servers with ATX supplies (which, coincidentally, I do), and wanted to make sure they came back on when power was lost (doh, didn't think of that, what are you running a server without a UPS for anyway???), then I might use a spare 386 or 486 I've got kicking around to do it.

Its a piece of cake to wire a relay to a parallel port, there's schematics all over the place on the net to do it. Trim the power supply's "on" pins, wire them to the relay, and you could easy control the power of 8 systems from the parallel port with maybe $40 worth of hardware. Its not exactly what the question was asking, but it might be useful. Maybe Slashdot needs something like this for when the machine misbehaves. ;)

Also useful in case of an errant init 0 instead of init 6...

Alternately, I think those Matrix Orbital displays with the keypad interface also have eight digital outputs on them, you could wire one of those to a serial port, and trigger the relays with those outputs. Running some monitoring software to put the status of the machines on the LCD, you've got a nifty monitoring/control system.

Re:OK, how about off?

Virtually all of those will hard power off even if the system is hung if you hold the power switch down for five seconds...

Sometimes they don't make that clear in the manual though. Its a problem if you have Linux on them and don't know that, and don't compile the kernel to shut the system off on halt.

Re:How can the MB control the power supply?

Here's what it will cost to run a computer all month for your electric bill at home if you pay $0.08 per kilowatt hour:

40 watts * 0.08 cents/1000watthour *24hours * 30 days
=$2.30

If you get the industrial rate at $0.02 per kilowatt hour, that will be about $0.58 per box.

UPS supplies seem to draw much more. They seem to have a type of saturation transformer where the iron core is intentionally saturated with excessive current to give a well regulated and safe voltage to your computer including isolating it from the mains.

Do it in hardware

Pin 14 on the ATX power supply connector is the "PS-ON" signal. From the ATX spec: [intel.com]

PS-ON is an active low signal that turns on all of the main power rails including 3.3V, 5V, -5V, 12V, and -12V power rails. When this signal is held high by the PC board or left open circuited, outputs of the power rails should not deliver current and should be held at a zero potential with respect to ground. Power should be delivered to the rails only if the PS-ON signal is held at ground potential. This signal should be held at +5VDC by a pull-up resistor internal to the power supply.

So, you should be able to hack it together to have that pin permanently grounded, so that the PS is always on. I haven't tried this; I think it should work fine, but don't blame me if your motherboard burns! :-)

There's also more info on ATX at http://www.teleport.com/~atx/ [teleport.com]

Do it with hardware

At hardware and auto parts stores you can get solderless connectors which tap a wire into an existing wire. You can use two of them to add a wire which connects the wires for pin 14 and a neighboring wire. They're usually blue plastic with a metal tab which you push into the two wires. Faster than soldering and can be removed for maintenance.

Problems with ATX

Because the power switch goes through the MB instead of to the power supply it makes it more difficult to narrow down problems with a computer. Is it the Power Supply? Is it the Motherboard? You can't tell without popping out the power supply and trying a known good one.

Additionally, is it just me or are the ATX power supplies more fragile than AT's? I've had to replace two of them myself in the period of one year, and my brother had to replace his a few months after he got his. I don't know about him, but I have a UPS (APC) so I know my power is clean. And I've been buying good quality power supplies. Argghh, I like the ATX case design thing-ma-bobs, but I REALLY don't like the power supplies. Is it just me? Am I cursed? Or have other's experienced my travails?

BTW, this is very recent so it's not because I'm using pre 2.01 compliant power supplies.

A couple options

There're a couple of options

Some ABIT motherboards can be strapped with a
jumper between the reset and power switch connectors to make them act in an 'always on'
fashion like AT motherboards do.

If you feel like experimenting a bit, you can
look in electronics parts catalogs for POST
transistors... these basically act like delayed
momentary-on relays. One lead goes to +5V,
the others go to the power-switch jumper.

--Z

Re:Mod to keep power on all of the time

Make sure you drop a 10kohm resistor in there, or you will reduce the life of the power supply...
We ran into this same issue with some boxes that we wanted to boot right up when power was applied, soldering gun and a handfull of Radio Shack Resistors later we were golden...

A

Re:Do it in hardware

We do this with all our servers (medium-size ISP in Europe) and it works without any problems at all.

Just take a bit of the isolation off the wire of Pin 14 (usually, bit NOT ALWAYS green). There's a GROUND signal on either side of this; take the isolation off that as well. Then solder together and wrap with insulating tape. Works like a charm. The only thing you shouldn't do is tell the mobo to power off; it gets a bit confused then.

This is where pin 14 is:
Looking down onto the contacts: (wires on rear)
v--(plastic tab)
===
+ + 0 # 0 + + + + +
+ + + + + + + + + +

# is pin 14. 0 is ground signal, should be a black wire.

My motherboard is ATX/Server, no problems here...

I have a Supermicro P6GDE (Dual PII 400, GX chipset) that has a jumper on the MB to choose, "Bios Power Stater, or Always on..."

As soon as I through the switch on the back of the powersupply its on, and if that switch is left in the on position then it comes on when the power returns after a black out, etc.

Note the case is a supermicro ATX750 or something(which is really an Antec Case and Powersupply) and it seems to be designed just for this...

Re:Example ATX power utilization

Initial note on credibility: I have been in the employ of the leading manufacturer of power supplies for more than a decade, in a highly technical role.

Your conclusions are accurate, but your actual data is wrong. The power is way too high to be wasting on every PC on the planet, but is in fact significantly lower than you measured. The power into a PC power supply cannot be estimated by measuring the AC current and multiplying times the known AC voltage. This would work with in-phase sine waves or DC, but not with the highly distorted current waveform present at the input of the power supply. Most of the measured current is in odd harmonics (3rd, 5th etc.) of the 60 Hz line. Multiplying a 180 hz current times a 60 Hz voltage will give alternating positive and negative power over time, with zero net power. (i.e. the 180 Hz power alternately flows from mains to computer, then from computer to mains). Your current meter measures this as part of the total current, though it doesn't reflect power. (Your local Utiliy's Wattmeter is not fooled - it reflects true power).

A Wattmeter is required, and the error can be 3x. (Actual power dissipated is lower is lower than you calculate.) This has been a real pain for a non obvious reason. Typically, we design equipment for use on a 10A service (assumed minimum size of fuse/circuit breaker; the ratings on wiring, wall sockets, etc.), which would imply that we could go up to 1200 W without problem even in a consumer environment. In fact, we will exceed 10A RMS on significantly lower power units. High power systems (750W and above) will almost always need an additional power stage, a Power Factor Correction (PFC) front end, which pulls power from the mains in a clean sinusoidal waveform to allow us to extract the maximum power at a given RMS current level.

Two upcoming related issues will increase the cost of power supplies: In Europe, there is already a new requirement on the books that will require that the level of harmonics in the current draw be reduces (same as saying that we will be required to make the input current look more sinusoidal). This will be a new requirement on low power (consumer sort of power level) equipment. Second, the "low power" energy savings modes will be made significantly more stringent, which will have significant impact not only on computers, but on all that consemer gear that stays alive waiting for your IR remote to tell it to power up. Even little wall-warts will be affected. the power drawn by old fashioned steel transformer based warts is on the order of a Watt or two - multiply that number times the number of cell phone chargers hanging off people's wall sockets 24 hours a day and you'll see staggering costs to society.