Min Power Supply

[ByteDown[H]ard]

I'm just curious what the min power supply needed would be if your running an all-in-one board with just a case fan, if its in a case and the Fan mounted on the HSF unit, unless its watercooled... I would think that the power requirements would be rather low, if all it does is sit in a corner and fold. If I can get away with 200w or lower?? PS rather than throwing a 400w PS in a dedicated folding box, it may reduce heat and electric bill just a little... I know there are tons of variables but I'm looking at getting rid of 5 duron boxen and replacing them with 2400 bartons on gigabyte all-in-one boards and if I can avoid a mini furnace on each board......

 

[Mattman]

Someone correct me if I'm wrong, but don't power supplies put out just the power they need at the time? I think a 400 is just capable of delivering more than a 200, but doesn't use any more if used with the same hardware. Corrections? Additions?

 

[Dark Ember]

Someone correct me if I'm wrong, but don't power supplies put out just the power they need at the time? I think a 400 is just capable of delivering more than a 200, but doesn't use any more if used with the same hardware. Corrections? Additions?

Yeah, thats right. Components only draw the current that they need. So, since they are operating at a constant voltage (unless you have a REALLY bad PSU), the power consumption is constantly changing. So, since you would have the same components no matter what the PSU, you would have the same power draw. The difference would be that your 400watt power supply would not be as taxed as your 200 watt power supply. So, the 400w PSU might actually run cooler than a 200w.

 

[Ch1m3r4]

I believe it would be running less efficiently though, which in theory would cause higher powerbill than a lower wattage psu.

edit: from the power supply faq

A larger PSU wont be as efficient if its not under enough load and it normally has fans with a higher CFM rating but also higher db rating which makes for a louder PSU. If nothing else your wasting your money when it could be put to use in other areas of the system.

of course by the same token, an underpowered psu is just as inefficient...

 

[BillR]

Not quite true.. I picked up a referb shuttle from the egg (sealed box) with a 200watt psu. It draws 1.8 amps at the wall. In it is a Barton at 2.0 gig, gig 0 ram and a 120gig drive and an ati aiw 8500 vid card.

Same exact configuration in a full case with a 400 watt psu draws 4.1 amps at the wall.

Tried a duron 850 in both just for fun, exact same results. SFF boxen cost more up front but do in fact cost less to run. Not to mention, they are so dang cute

Fold on

BillR

 

[ByteDown[H]ard]

hummm, goes to show how much/little I know about one of the major components of the systems I've been building for 12 years or so.. live & learn and fold some more...

 

[Mattman]

ByteDown[H]ard, don't fret...this isn't an area that most people likely to get into. You only wonder about these types of things if you're running a house full of folding machines or planning electric circuits for large server rooms.


BillR, I'd be curious to know if your PSU comparison was an apples to apples one. How's the quality of the 400W PSU? Does it have multiple, big fans? I could understand some inefficiency in the heftier PSU, but that seems like a really big difference. If that's accurate, then we need to come up with some reliable, low Watt PSU's to recommend for basic folding machines so we can stop wasting so much juice and generating so much heat!

 

[Hyperion]

Same exact configuration in a full case with a 400 watt psu draws 4.1 amps at the wall.

Thanks for confirming my suspicions BillR ... good post there!

 

[CIWS]

I'm just curious what the min power supply needed would be if your running an all-in-one board with just a case fan, if its in a case

Several small systems run on power supplies starting at around 100watts and it's not uncommon to find MicroATX systems with 150-180 watt supplies in them.
Although this newest Antec Aria case for MicroATX I just bought comes with a 300watt supply, but I wanted that in order to run a bit "heavier" SFF system with.

 

[BillR]

Very legitimate question Mattman

I did the same study 2 years ago with very similar results. I use a high quality mulitmeter, cut a computer power cord in half and hooked one half in series with the AMP meter section of the meter.

Using quite a number of different PSU combos the measurements were all very consistent. My understanding of the different PSU’s is the higher the quality the less chance of “SAG” or voltage drop when the power lines go funky. Whether it was a $7.95 400-watt unit or a $100 400 watt unit the overall current draw was very close. Within a quarter AMP. Number of PSU fans doesn’t seem to matter much.

I had tried using the “AMP Probe” type meter but it was far too generous. I only did this after noticing I had taken out a 20 AMP breaker with 8 Boxen. Not to mention my wife pointing out that our electric bill had not dropped since we turned off the AC units. (I had just added 4 boxes at that point)

Also of note, it doesn’t seem to matter if your at idle or 100% cpu usage, same current draw at the wall.

The design of today’s PSU is to make it non-dependent on the AC source for additional power needs from inside the computer, as such it makes all it’s power all the time.

I by no means am trying to be the final word on this; I’d gladly welcome any suggestions on lowering my now horrible electric bill

BillR

 

[pageian]

Interesting stuff.

BDH, I can tell you that I run several all-in-one boxen on 180 psu's with no problems. They came stock in the Enlight mATX cases they're in so I believe they are descent quality. I've been running a few of the for a couple years now with no problems.

Wish I'd read this before I ordered that 250 psu from the Egg the other day.

 

[gnewbury]

Very legitimate question Mattman

I did the same study 2 years ago with very similar results. I use a high quality mulitmeter, cut a computer power cord in half and hooked one half in series with the AMP meter section of the meter.
<snip>
Also of note, it doesn’t seem to matter if your at idle or 100% cpu usage, same current draw at the wall.

The design of today’s PSU is to make it non-dependent on the AC source for additional power needs from inside the computer, as such it makes all it’s power all the time.

I by no means am trying to be the final word on this; I’d gladly welcome any suggestions on lowering my now horrible electric bill

BillR

BillR -
Not being "electrical smart" I'm not sure watt you mean by "current draw at the wall".
My system - dual gateway server w/ 2 PIII 933's, several hard drives, Viewsonic 20G uses
juice, but not like an AMD.
According to my trusty Kill-A-Watt, and only a few readings, usually for about a minute each -
Monitor off (this made it difficult to type ):
1% CPU usage (last I looked) - 1.12 AMPS, 85 WATTS
100% CPU usage (2 CPU's folding) 1.65 Amps. 128 WATTS, 122.4 volts

Monitor back on (whew!)
1% CPU usage (last I looked) - 2.75 AMPS, 205 WATTS
100% CPU usage (2 CPU's folding) 2.95 Amps. 245 WATTS, 122.4 volts

The proteins are a p262 and a p266.

Thus it seems that when both CPU's are folding more juice flows out of the wall, about 40 watts..

All readings were taken after the system warmed back up for a few minutes.

What I think this proves is that I need to get a 20 inch LCD monitor so I can save on electricity
Plus it's darned hard to type with the monitor off.

/edit - all running through an UPS

 

[dwdawg]

I have two MATX all-in-one boards running Barton 2500's (not mobile) at 2.3 Ghz on 150W cheap ass power supplies I bought on Ebay for about $9 ea.

Just call me lucky

 

[Localan]

Someone correct me if I'm wrong, but don't power supplies put out just the power they need at the time? I think a 400 is just capable of delivering more than a 200, but doesn't use any more if used with the same hardware. Corrections? Additions?

Correct. Think of this like a Ferrari (or '69 Camaro if you like that better) with 400+ horsepower. It will require less than that 400 hp to maintain 65 on the freeway even though you could use all 400 to try to outrun the police officer you see in the rearview mirror.

Not quite true.. I picked up a referb shuttle from the egg (sealed box) with a 200watt psu. It draws 1.8 amps at the wall. In it is a Barton at 2.0 gig, gig 0 ram and a 120gig drive and an ati aiw 8500 vid card.

Same exact configuration in a full case with a 400 watt psu draws 4.1 amps at the wall.

BillR,
Not going to call you a liar, but you might want to check exactly what you were measuring again. Watts = volts x amps, so in the case of your Shuttle you would have: Watts = 1.8A x 120V (<-- this typically will be somewhere between 115 and 125 inside your home) = ~215 Watts, which is very believable due to fluctuations of efficiencies of the power supply and voltage at the wall as well as the number of components in the system.

However, the "full case" example you cited would be: Watts = 4.1A x 120V = ~475 watts. While it is believable (although unlikely) that the power supply could handle this much during a spike, I don't think that it could for a continuous period.

Anyway, like in the car example above, the power supply should only deliver what is asked for by the computer. This number might fluctuate more at startup as the PSU (<---- No, not Penn State University) is busy trying to charge all of the capacitors and whatnot on the board and within the PSU itself. Monitoring this through a UPS or the like will show you that the power usage will go up if the processor/video card/CD burners/etc. are working harder or at all. Or down for the reverse situation.

BTW, if your 8 computers on the same circuit all used this same 4 amps each, then that 20A circuit breaker, which is only rated to continuously handle 16A by the National Electrical Code, would have been trying to carry 32A. No wonder you had issues.

Anyway, I didn't mean this as an attack on your measuring skills, it's just that this question comes up every once in while, and the Electrical Engineer in me gets released for the world to see. I promise that I will go back to my cave for awhile and try to figure out what I need to do to hold off this rash of young folders trying to knock me out of the top 25 all of a sudden.

 

[gnewbury]

<snip> Watts = volts x amps, so in the case of your Shuttle you would have: Watts = 1.8A x 120V (<-- this typically will be somewhere between 115 and 125 inside your home) = ~215 Watts, which is very believable due to fluctuations of efficiencies of the power supply and voltage at the wall as well as the number of components in the system.
<snip>

I thought that was the formula, (VAW, or Virginia Warriors, or V*A=W) but why am I getting numbers in the 75 to 80 volt range if I Excel the numbers I get from my Kill-A-Watt?

 

[BillR]

All very good points. If you look at the side of your PSU you will see two ratings. One at 110~127 volts and one at 220 volts. The PSU in front of me at the moment says:

Voltage 110~127 @ 8A
Voltage 220 @ 4A

This is pretty normal for a PSU in this range. Although it doesn't draw it's full 8 amps at idle with no load, it does draw 4 AMPS when hooked to a computer.

My power company finally admitted that their "estimates" of consumption were based on a "Typical home computer in 1995, used an hour or two a day". They took their readings on my machines and they were within 1% of my readings.

Localan, they too (The power company) confirmed, 3 to four computers PER circut rated at 20 AMPS, at the current rate of usage was safe. I have since split the load over 2 30AMP circuits, each circuit on a different side of the breaker box for correct balance. The power company is stateing they are going to re-evaluate "The cost of running a computer". I don't believe that for a moment, but we will see.

All this came about when my electric meter was replaced last year but a new "Better" one that they can read over the power lines. The Better one raised my bill a whopping 30%. It was then I started taking readings with a real meter.

Anywho, that's been my experience to date..I welcome all comments and theories, I'm a Tech, not an engineer

Fold on folk

BillR

 

[Vertigo Acid]

From my experience planning and running LAN parties, I can tell you that you can easily throw 50 people on (4) 20 amp circuits and be fine. Your reading still bothers me, however. If they were totally accurate, that would mean ~200w were being lost to inefficiency. Switched mode power supplies do not have the means to load themselves, so all of the power was ending up as heat. Odd. And low quality. I'm now tempted to do a similar experiment between my three good power supplies, a 300w Enermax, a 480w Antec True, and a 520w Vantec Stealth, to see how they differ.

 

[Little_Kenny]

I'm now tempted to do a similar experiment between my three good power supplies, a 300w Enermax, a 480w Antec True, and a 520w Vantec Stealth, to see how they differ.

Please do. More info is a good thing.

 

[Localan]

All this came about when my electric meter was replaced last year but a new "Better" one that they can read over the power lines. The Better one raised my bill a whopping 30%.

You need to go back and find that letter again and check to see if it looks like this: "We will be replacing your existing meter with a better for our profit margin meter".

More seriously though, that is definitely an issue if your bill went up 30% for no reason other than a meter change. I would almost say that you should have them check the CTs in the meterbox to make sure that they have the proper ones (turns ratio, etc.) for your panel, but that would not explain the abnormally high reading inside the house, nor the fact that your measurements indicate no power usage difference between idle and full-blown on the computer itself.

As a point of reference, my HTPC (mobile XP2400+, motherboard, video capture card, 16x DVD, 2x200GB HD, 1x160GB HD, AIW 8500DV, SB Audigy) reads a difference of about 50 Watts (275-225) from idle usage to as-busy-as-I-can-get-it.

 

[BillR]

You need to go back and find that letter again and check to see if it looks like this: "We will be replacing your existing meter with a better for our profit margin meter".

More seriously though, that is definitely an issue if your bill went up 30% for no reason other than a meter change. I would almost say that you should have them check the CTs in the meterbox to make sure that they have the proper ones (turns ratio, etc.) for your panel, but that would not explain the abnormally high reading inside the house, nor the fact that your measurements indicate no power usage difference between idle and full-blown on the computer itself.

As a point of reference, my HTPC (mobile XP2400+, motherboard, video capture card, 16x DVD, 2x200GB HD, 1x160GB HD, AIW 8500DV, SB Audigy) reads a difference of about 50 Watts (275-225) from idle usage to as-busy-as-I-can-get-it.

Right now it's a matter of contention between my electric company and the PUC. I want proper readings done with a recording graph. I agree, I'm being screwed

My readings were taking with a Simpson Digital Amp-meter in series with the power cord, as such I didn't convert that to watts, just a straight current draw reading. What also drew all this to my attention was a Belkin Surge protector. After about 6 months of use with 6 computers two machines would shut down and not be restartable. The unit is rated at 15 AMPs and 1500 watts continious. Took it apart, found that the "rails" inside were all burned and warped. Switched to a Monster unit, no problems, but that is what caused me to start taking my initial readings.

Anywho, that's my story...

BillR

 

[privatteer]

Electronic equipment (switching power supplies) draws current differently than non-electronic equipment. Instead of a load having a constant impedance drawing current in proportion to the sinusoidal voltage, electronic devices change their impedance by switching on and off near the peak of the voltage waveform.
The circuit of the power supply only draws current from the AC line during the peaks of the voltage waveform, thereby charging a capacitor to the Peak of the line voltage. The DC equipment requirements are fed from this capacitor and, as a result, the current waveform becomes distorted.
Switching loads on and off during part of the waveform results in short, abrupt, nonsinusoidal current pulses during a controlled portion of the incoming peak voltage waveform. These abrupt pulsating current pulses introduce reflective currents (harmonics) back into the power distribution system. These currents operate at frequencies other than the fundamental 60 Hz.

The power supply in a modern personal computer is a good example of a non-linear load.

Average response ammeters are only accurate when measuring 60 Hz loads that have sinusoidal current waveforms and cannot accurately measure the current of nonlinear loads. The reason is that nonlinear loads draw current in a nonsinusoidal manner, which produces reflective harmonic currents that operate above 60 Hz; both of these conditions are beyond the meter's design criteria. When an average response ammeter is used to measure nonlinear load current, the results can be inaccurate readings of as much as 25% to 50% the actual true-RMS current. As a result, the actual current of a circuit can exceed the rating of conductors and equipment. The actual current cannot be detected with the average-responding ammeter!

When induction-disc watt-hour meters (Older power meters) are monitoring non linear loads, depending on the content of the harmonics, the disk may run slower or faster, resulting in erroneous readings. Many power companies are rolling out new True-RMS power meters to correct this.

In order to perform basic electrical trouble shooting for non-linear loads you must have an ammeter that provides true-RMS and instantaneous peak current ratings of the circuit. This meter must have the capacity of measuring the electrical characteristics of the waveform by sampling many points along the waveform. True-RMS meters are designed for just that, and they are accurate for both simple (sinusoidal) and complex (nonsinusoidal) alternating and direct current waveforms. Average response meters are only accurate with simple sinusoidal alternating current waveforms, ie lights, basic motors; not the complex waveforms resulting from nonlinear loads.

To say it bluntly, if your useing an average responding ammeter/ wattsmeter to get your readings you might as well make a lamp out of it because it is useless!

When a current, or voltage, operates at other than the fundamental 60 Hz frequency it is said to operate at a specific harmonic order (3rd harmonics operate at 180 Hz; 5th harmonics operate at 300 Hz).

Harmonic currents sometimes cause false circuit breaker tripping. Peak sensing circuit breakers often will trip even though the amperage value has not been exceeded. Harmonic current Peak values can be many times higher than sinusoidal waveforms.

Because reflective harmonic currents operate at frequencies higher than the fundamental, their effect in the electrical distribution system can be considerble.
Odd number harmonics ( 3rd, 5th, 7th, etc. ) are of the greatest concern.
Even number harmonics are usually mitigated because the harmonics swing equally in both the positive and negative direction.
The heating effect causes the greatest problem in electrical distribution systems and equipment. Electrical equipment often overheats and fails even when operating well below the design ratings. The increase in temperature is directly related to the increase in RMS current and the increased inductive heating effects of eddy currents, skin effect, and hysteresis. (Increased eddy currents as transformers,etc designed to cancel the 60hz frequency best)

On a note at Lan parties, which I help run serveral large events.
On balanced, three phase systems with no harmonic content, the line currents are 120° out of phase, cancel each other and result in very little neutral current. However, at lans there is distortion on all of the phase currents and almost 100% non-linear loads, the harmonic currents increase and the cancellation effect is lessened. The usual result is the neutral current is significantly higher than realised. The triplen harmonics (odd multiples of three) are additive in the neutral and can quickly cause dangerous overheating.

In theory, the maximum current that the neutral will carry is 1.73 times the phase current and if not sized correctly, CONSIDERBLE overheating will result.
A neutral burning off at a plug, or board can result in significant damaged equipment and has resulted in several dead pc's at one of my earlier events.
Once about 200amps per phase current is reached however the neutral drops to about 1 times phase current as some canceling of the 3rd harmonics takes place, thus only needs to be considered for the sub-boards.

Note: Written for 240v,60hz ac supply, but equaly applies to other's.

 

[Little_Kenny]

Cool. Thanks privatteer. Good reading.


Fold On!