Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- Question about PSU's
May 2, 2009, 11:21 am
rate this thread
I bought a new bare-bones system some two months ago, and have blown
two new 400W PSU's now. I think I need to go for a stronger PSU.
I notice my M3A78-CM motherboard has a larger so-called 4PX20
connection plus another separate 4P square connection.
I pulled one of my old PSU's out of the close that has just a 20pin
connector and the separate 4P. I connected it to the motherboard and
things are running fine. So I guess I can go with this until I get a
But I have two questions.
1) What calamity am I apt to suffer because I have left the one 4P
connector (on the 4PX20) open?
2) What spec do I look for as I shop for a replacement stronger (say,
600W) PSU? I want to be sure the new PSU has both the 4P connectors.
Re: Question about PSU's
When I look at the picture here, I see a 24 pin main connector, and a
2x2 ATX12V connector for processor power.
The four pins on the end of a 24 pin connector, have four
different colored wires. The wires carry redundant 3.3V, 5V, 12V1,
What purpose do redundant wires serve ? They increase the ampacity of the
interconnect. Each pin of the connector, can safely handle 6 amps of
current flow (the exact number may depend on the gauge of wire leading
to the pin). There are five red wires in the 24 pin wiring harness,
which all carry +5V. 5 wires times 6 amps each, gives 30 amps maximum
that the motherboard could draw, without burning the connector. In
practice, a motherboard designer would likely try to avoid going
all the way to 30A, since the current sharing in a motherboard
isn't perfect. (Some pins carry more current than others. One
motherboard I checked, the spread is +/- 1 amp.)
The most important pin of the new ones, is the extra 12V1 pin. Since
the 24 pin connector now has two wires carrying 12V1, up to 12A can
be drawn from that rail. On your motherboard, there is one video card
slot, and the most slot power I've heard of in a video card slot,
was a load of 4.35A. So 4.35 amps doesn't exceed the single wire on
a 20 pin connector. There could be an issue if two video cards were
present, as 2*4.35 = 8.7A, and then you would benefit from the use of
a 24 pin connector.
In terms of power supply shorting, there are a couple cases to consider.
The motherboard could have a (partial) short, between a rail and ground.
That would strain the power supply, and bring it close to its limits,
causing it to heat up internally. Or, if the overcurrent detection in the supply
was actually working, it might shut off the supply, at about 30% over
A second kind of fault, would be rail to rail. Say the +5V rail on the
motherboard, had a partial short to 3.3V. Current from +5V would flow
into the 3.3V rail. The ATX power supply is not "push-pull" and has
no way to compensate for that. The 3.3V rail voltage will rise, the
feedback system will hit the wall, the supply will make its maximum
possible correction to the disturbance, but the 3.3V will still be
rising. If the power supply is well equipped, it will have OVP. The
supply will shut off, if the 3.3V rises high enough. If it doesn't
have OVP, then perhaps the motherboard components will be fried
by the out of spec voltage.
When a motherboard pops two supplies, either the calculation of
required power is wrong, or the system is drawing more power than
it should be. To check that, you can use a clamp-on DC ammeter.
(I have the 380947, in this document.)
To use it, I place the five red wires in the ATX power supply
bundle, all within the jaws of the meter. Set the meter to 40A DC
max range, then take a measurement. The meter sums the magnetic
field around the five wires, so is able to measure the combined
current flow. In that way, I can determine whether the loading
is "normal" or not.
If I take the meter, and clamp it around an AC cord, it reads zero,
because the magnetic field around the two power carrying conductors
cancels. To measure AC power, you need a "cheater" cord, where the
outer insulation of the cord is removed, leaving the three insulated
wires inside loose. By grabbing the hot wire, and clamping around
that, it is possible to make an AC measurement. I have a couple
cheater cords I made myself, as I'm too cheap to buy one from the
I've also used that meter to work on the car. When my car was having
trouble starting on a cold morning, I used that meter to measure the
peak current consumed, and it measured 150A on the 400A DC scale. So the
meter can be used for more than just motherboards. I also use it
for checking the current draw on the central air conditioner. (The
fan on my central air, draws twice the rated current, and some day,
it is going to burn out.)
Is it worth investing money in a specialized meter like that ?
Probably not. But it does have the advantage, of allowing a
"non-contact" method of measuring current. No fuses to pop, no
accidents. Inside the device, is a Hall probe, which is what
converts the magnetic field into a voltage. The jaw design
is a flux concentrator, that helps the device only measure
the magnetic field encircled by the jaws. Fields outside
the jaws are ignored.
To make a current measurement with a conventional multimeter,
1) Measure individual wires, one at a time, due to the 10 amp
limit of the highest range on a cheap multimeter.
2) Purchase a 24 pin to 24 pin extension cable, and chop each
wire in two, then insert the multimeter between the two
ends, then turn on the computer and make a measurement,
while the computer is running your favorite test case
So it is possible to make some measurements, but by measuring
the individual wires, there could be additional measurement
error. And with that many wires, I'd probably buy crimp terminals
and a terminal block, to make it easier to manage all the
cut ends of the wires.
To measure currents over 10 amps, you can make your own shunt
out of manganin wire. You have to set the shunt resistance low
enough, such that the manganin doesn't get too hot. Manganin
is selected for its low temperature coefficient of resistance,
which is why it would be preferably to using some nichrome wire
scavenged from your toaster.
A Hall probe based meter, avoids some of those issues, and
makes it much quicker to measure system power consumption
on a per rail basis.
Re: Question about PSU's
On Sat, 02 May 2009 07:21:11 -0400, firstname.lastname@example.org wrote:
Are you using the integrated video, or at least a
lower-tiered video card instead of a high end gaming video
If so, a high quality 200W PSU should be enough. Seriously,
using a board with all-integrated features and a hard drive
or two plus an optical will typically peak at under 100W
total, perhaps a bit more with overclocking and/or a higher
powered Phenom, but this is the ballpark you're in,
definitely not needing over 400W PSU.
What is more likely is your PSU is a generic with not only
false ratings but also substandard parts and/or omission of
parts that contribute to longevity, like inrush current
limiting. This latter detail is not necessary to focus on,
any decent PSU will have it.
If your old PSU has a 5V rail to 12V rail current ratio of
at least equal or lower than 2:1, in other words suppose it
is rated for 28A @ 5V and 14A or more at 12V, it is probably
going to be fine powering the system long term, providing it
is not an overrated generic and that it doesn't have enough
years runtime on it yet that it's near the end of it's
expected lifespan. This is assuming you bought a board with
integrated video to make use of that video, that it is a
typical/basic system and not a gaming rig or a fileserver
with nearer half a dozen hard drives in it, or any other
atypical parts that use a lot of power.
None usually, a conservative 6A per connector pin estimate
and two pins for 12V and two for ground will allow 12A *
12V, 144W which you can multiply by the typical CPU VRM
power circuit efficiency of roughly 93% to arrive at 0.93*
144 = 134. So long as your CPU's max rated power is 134W or
less, and/or if you overclock it you stay below 134W, and
assuming the power supply contacts and mating socket are in
good condition, there is no need to use more than a 2x2, 4
pin plug with that.
Given what I wrote above, you probably don't need anything
special, a typical high quality 350W PSU with a 2:1 or lower
ratio of 5V:12V current should suffice. Buying a 600W or
higher PSU could actually be detrimental, in not only
costing more but being less efficient (all else being equal)
with a system that doesn't need much power, just a PSU
designed well enough to not self-destruct.
Hopefully you can get some money back as a refund or credit
towards a better PSU than what you already had. Might
depend on what your time is worth, even if a seller balks at
this in some countries, or with some credit cards, or
exploring warranties and company customer ratings it can
eventually get you a new PSU if you put in enough time.
However, if the original PSU was a 400W generic in a
barebones bundle, it was probably accounting for about $15
to $20 worth of the cost, it would be easy to spend more
time than it's worth.
Re: Question about PSU's
I probably misread what you wrote, ignore the above reply
but the situation is similar, that a typical system with the
board you have will not need the extra current a 24 pin plug
offers over a 20 pin plug. The extra 4 pins are redundant
to decrease resistance through more supply lines and
connector pins for systems that need more than average
current while a system with typical components and an
efficient motherboard like yours will use less current than
a typical system that runs fine off a 20 pin connected PSU.
- » Newest MemTest86, MemTest86+ incompatible with Sandy, Ivy chipsets?
- — Next thread in » Computer Hardware