PSU questions

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View

Hello all,

My computing style is modest in terms of range of use and power
requirements, and as such when it has come time to replace my PSU I want
one which is modest in it's power rating so as to save both power, money
and in line with green ethics.

Does anyone therefore know where I can get a full ATX size PSU rated at
120 OR 150 OR 200W?? Must be Full (m)ATX size and not micro/pico/mITX etc.

Passively cooled would be a VERY nice touch too!

Awaiting your wisdom.


Re: PSU questions

Keiron wrote:
Quoted text here. Click to load it

The power consumed is *not* determined by the overall rating.
That isn't what you should be sorting them by.

You overall power bill, is proportional to the load presented
by the computer. As an example.

Power Supply = 1000W. Internal load  200W. Power used = 200W
Power Supply = 1000W. Internal load  500W. Power used = 500W
Power Supply = 1000W. Internal load 1100W. Power supply shuts off on overload.

Power is only drawn from the wall, in proportion to the load.
The rating of the supply (1000W in my example), determine the
upper limit for the internal load. I tried to draw 1100W from
the 1000W supply, and it overheated and shut off.

The power consumed from the wall, consists of two component parts.

      internal_load + waste_heat_from_PSU

A standard ATX supply is 68% efficient. A high efficiency can be
80% efficient or maybe 85% efficient.

Two examples. We'll draw 200W from a 68% efficient and an 85% efficient.

   ------ minus 200W = 294.11 - 200 = 94.11W waste heat.

   ------ minus 200W = 235.29 - 200 = 35.29W waste heat

The 85% efficient supply runs cooler than the 68% efficient one.
With the same internal load (200W), one PSU kicks out 35W of heat
and the other kicks out 94W of heat. In terms of the power
bill, the inefficient supply draws 294W while running the computer,
while the efficient one draws 235W. You can figure out the difference,
multiply by the dollars per kilowatt hour, and figure out the
payback period for usage of the more efficient supply. Efficient
supplies have a higher purchase price.

To summarize

1) Buy a supply with sufficient overall rating, to handle the load.
    A 500W supply would be good for a 300W computer. Leave a little
    slack space for stability and long life.

2) Buy a high efficiency supply, to save on the component part of
    waste heat emitted by the power supply. Power supply efficiency
    is not a constant, but varies slightly with the load. So the
    85% efficiency rating, might be at 200W load. The efficiency
    might be a little lower, if the total load was smaller.

Since you gave no idea about the hardware inventory inside the
computer, I cannot pick an appropriate model for you.

As an example, this one claims to be 90% efficient. Cost is $170 USD,
so there would be a "payback period", to make back the money in power
savings. It can handle a load up to 750W, but could just as easily
handle a computer internal loading of 100W.

CORSAIR CMPSU-750HX 750W  $170

This is the efficiency curve. The unit is 89% efficient at 150W loading,
so is still good even when powering an energy saving computer.


Re: PSU questions

On Sat, 27 Jun 2009 15:55:18 -0400, Paul wrote:

Quoted text here. Click to load it

Hi Paul,

Thanks for your reply. I understand that a the maximum rating of any
given psu isn't continuously drawn by a system however I was working on
the reasoned, but unscientifically proven, assumption that a modern power
supply with a max rating of between 100-200W would be more efficient
across the range of loads drawn from it. This I thought would be
particularly efficient given the relatively low requirements of my system
(celeron 1ghz; mobo with onboard sound,video etc; cd drive, floppy disk
drive and that's the lot).

Also and again a reasoned but unproven assumption I made was even at say
90% load on the lower rated psu it would produce less heat. I know this
doesn't necessarily make logical sense tho.

Also, and perhaps wishful rather than anything else, I was hoping 200W
would be cheaper than it's 500+W counterpart.

Do you know if either of my assumptions hold true? I'm struggling to
prove it either way being as I can't seem to locate and 'modern' 200W

Incidentally, do you know if components attached to the power supply but
which are not constantly in use, CDROM and floppy drives for example,
create load on the psu when not active??


p.s. awesome looking recommendations but beyond my pocket and well too
sexy for my rig.

Re: PSU questions

Somewhere on teh intarwebs Keiron wrote:
Quoted text here. Click to load it

If there were such beast it'd be likely to be more expensive rather than
cheaper simply due to supply / demand dynamics. They might sell one 200W per
thousand 500W boxes so they'd have to be correspondingly more expensive.

Then there's the fact that, if the *were* cheaper it'd be because of cheaper
components. A decent 500W PSU, being used as a 200W unit should last 2.5
times longer (or more) dur to less stress on components.

[rest shipped]

"Build a man a fire and he'll be warm for a day. Set a man on fire and he'll
be warm for the rest of his life." Terry Pratchet, 'Jingo'.

Re: PSU questions

Keiron wrote:

Quoted text here. Click to load it

Power supply development, follows where the market is. Nobody would waste
an extra moment, improving a 250W supply.

You can probably still buy low power rating supplies. (After all,
there have to be replacements for all those blown Bestec 250W
supplies.) But those kind of supplies will be 68% efficient,
so you're not any further ahead. When the efficiency is not stated,
it is a standard, poor number like that. It doesn't matter whether
it is a 200W 68% efficient or a 1000W 68% efficient, it is still
going to be inefficient and kick out the waste heat in proportion
to the computer internal load.

Simply go to your favorite retailer, and look through the "80+ efficient"
supplies, for something you can afford. I picked the $170 one, to
show an example of the *maximum* possible power savings. There are
slightly less efficient ones, still not cheap, that you could buy.

And yes, components not doing anything, can use power. I own a
clamp-on DC ammeter, and just checked these two devices in the
computer I'm typing this on. There is no media present in either

CDRW drive  5V @ 0.37A
            12V @ 0

Floppy      5V @ 0
            12V @ 0

And that says, my CDRW optical drive, wastes power via its controller
board, even when it isn't doing anything. There is a processor on
the controller board, and that is what sucks up the power while
idle. So I'm wasting 1.85W for nothing right now.

Even if your hard drive, were to spin down when idle, the controller
board on that also uses +5V. I would expect a similar result
from a spun down hard drive. (Hard drive spin state, can be controlled
by the Power control panel in your OS.) A hard drive is rated for a
minimum of 50000 start/stop cycles, so that tells you how many
times it is safe to spin down and spin up the hard drive again.
That would be a trade off on wear and tear on the hard drive,
versus the power saved.

If you place your optical drive, in a 5.25" enclosure, and switch it
off at the back, that would save the 1.85W.

For maximal power savings, as well as switching off at the back of the
enclosure, you'd have to unplug the wall wart as well. I don't know
how efficient the wall wart is, when no power is drawn by the enclosure.
(I have one of these next to me, and it has my DVD-RW in it. It is
currently switched off at the back, and the USB ports cannot see it.)

It is actually pretty tricky, to make the right decisions in each case,
in terms of power savings. You could go to a lot of trouble, believing
a certain idea is going to save power, only to forget some little detail.
It isn't as simple a task as you might think.


Re: PSU questions

On Sat, 27 Jun 2009 17:25:17 -0400, Paul wrote:

Quoted text here. Click to load it

Annoyed by the fact that 'the common sense' would suggest more power is
development I sadly agree. Strikes me that many don't use there PCs to
their full power requirements, but hey, I'm on no crusade.

Quoted text here. Click to load it

Cheers. I'll almost certainly payout for an efficient one, seems like a
winner all round.

Quoted text here. Click to load it

Thanks for this research. Given I can boot from USB I may well renew
consideration for an installed optical drive, and I SD card harddisk.
Should be a huge heat and noise bonus too.
Quoted text here. Click to load it

Incontestable. While I'd like to go on using older hardware till the end
of its life it's becoming apparent that newer hardware and software
support are much more energy effective. This all in one(cpu included) ITX
board for example uses only 40W (customer calculations but probably not
far off if at all) fully loaded:

At £62 seems like good financial sense too.

Thanks for your advice, very helpful.


Re: PSU questions

Keiron wrote:

Quoted text here. Click to load it

That one has the 330 dual core, and that to me would be the
minimum. I wouldn't want a single core version. I've already
retired an AthlonXP 3200+ because of its single core and
slight hesitation.

There are plenty of neat toys out there. For example, this
has an Nvidia IGP, the 9400M. The board reviewed, uses a 330. /

On video playback, it draws 27W.

Lots of fun to play with, as long as the price is right.

There is a suggestion here, the intro price might be $189 USD.
It includes some kind of Wifi card. It also comes with an
external power brick (so the DC to DC power conversion might
be a limitation).

For non-video purposes, your choice is a better one when
the price is taken into consideration. If you were building
an HTPC, the Ion version might be a better choice.


Re: PSU questions

Somewhere on teh intarwebs Paul wrote:
Quoted text here. Click to load it

I've looked at that one before (or it's Atom 330 predecessor) and the thing
that bugs me the most about them is the '533' 133MHz bus. My five year old
laptop was recently replaced as my main machine because of the 133MHz
bottleneck. For a while I was going to use one of these ITX boards instead.
All I can say is that Intel are deliberately crippling these boards so as to
not cut into their bigger markets. I can't see why they couldn't have a
200MHz FSB (at least). Modern laptops have up to 333MHz FSB so it'd
certainly be possible to do with an ITX board.

Paul, I have a friend with one of these and we've done CPU-only benchmarks
and the Atom 330 comes out at 111 marks x 2 = 222 marks while the Barton @
2.2GHz comes out at 208 marks. Subjectively he says the Barton seems more
responsive, probably due to the 200MHz FSB. Swings and roundabouts. The
Barton does suck around 5 x the power...


"Build a man a fire and he'll be warm for a day. Set a man on fire and he'll
be warm for the rest of his life." Terry Pratchet, 'Jingo'.

Quoted text here. Click to load it

Re: PSU questions

~misfit~ wrote:

Quoted text here. Click to load it

That is pretty impressive for a little processor like that.

The S462 is a DDR FSB, so 200 * 2 * 8 bytes per transfer is 3200MB/sec.
If the Atom has an 8 byte interface, and runs at quad pumped 533, that is
So it should be a slight bit better than the Barton.

As far as I'm concerned, the FSB on S462 was the bottleneck. It
made no sense, on my motherboard at least, to have 6400MB/sec
memory bandwidth, and a 3200MB/sec FSB.

On my Barton at 2.2GHz, I measured 80 watts going into the
5V rail on the motherboard (clamp on DC ammeter around all the
5V wires at the same time). But I don't know whether Vcore
is the only load on that rail or not. (That motherboard uses
+5V for the processor. The memory might have been running
from +3.3V.) The 80 watts was while running four copies of Prime95.

I dialed that box down to half the speed, but I couldn't stand
it for very long :-) It made a slide show out of my favorite


Re: PSU questions

On Sat, 27 Jun 2009 18:28:18 GMT, Keiron

Quoted text here. Click to load it

There are two primary problems.

That era of system uses primarily 5V current, since that is
the rail a Celeron 1GHz or so would use for it's VRM supply
circuit on the motherboard.   Older ATX spec PSU designed
for more 5V current (versus 12V) were designed before
systems used as much power as they "might"  now and before
many people were concerned much about efficiency, so unlike
tweaking supply capacity to load to get over 85% efficiency
with today's systems, you'll probably end up with something
closer to 65 to 75% efficiency.

Generally speaking on the older designs if it is passively
cooled, they simply took a PSU that would've been capable of
higher load, a higher capacity PSU, and derated it for the
lower wattage so it could run passively.  In other words, a
PSU with a fan will be what you want and it will last
longer, be more efficient all else being equal, and since
the system doesn't use a lot of power the fan can run at
pretty low RPM so long as you have a rear case exhaust fan
and ample case front intake area.

If the system has no exhaust except for the PSU, and no
fairly good airflow rate intake fan, your PSU fan will need
be all that much faster.

Generally speaking, you face another problem in that some of
the PSU optimized for more 5V current may be fairly old
stock and the capacitors have shelf-rot, have less remaining
lifespan due to this.  For that reason I would avoid buying
the PSU from any place that might sell surplus or old stock,
would not buy from ebay either, only from a retailer known
to  turn over stock fairly quickly if you can't get a
guarantee about the manufacture date.

If you really want fanless, here is one option.

Quite pricey for what it is.

Here is what I feel is more appropriate for your system,
although it may still use a mediocre quality sleeve bearing
exhaust fan that would be replaced for longest lifespan... a
replacement fan would be 80mm x 25mm thick and rated for
roughly 2000 RPM to keep noise low.

Unlike with modern PSU, many of the older OEM PSU rated for
90W to 200W weren't more efficient than the 250W of the same
era... basically the main difference was just that the
transformer was larger.

Site Timeline