David said:
Especially as there are other devices pulling +12V power.
Heh, the GeForce 9600 GT requires 26. That's why I recognized the
problem. Had a power supply that was rated for 24 amps on the
+12V rail. Had spontaneous reboots and shutdowns, with no log
messages, or obvious causes. Figured out it was the ps when adding
a third usb device forced an immediate shutdown.
Wouldn't boot with three usb devices connected. Any two of the three,
was ok, but not all three.
Replaced the ps with a 750 watt unit with 52 amps on the +12V rail,
and it's been working fine since then.
Regards, Dave Hodgins
OMG. The legend of Zelda lives on :-(
Whacking great power supplies are not required.
Intelligence is required !
Think about it. Examine the evidence.
"Wouldn't boot with three usb devices connected.
Any two of the three, was ok, but not all three."
On modern systems, USB bus power comes from the +5VSB rail.
The power supply label may list
+5VSB @ 2A
or it might show
+5VSB @ 3A
When you have a problem with the system falling over, and
the problem seems to be related to the number of USB bus
loads, then the problem is hinting at the +5VSB rating,
not the +12V rating. A 750W supply, has a *huge* 12V rating,
but none of that huge rating is of any consequence, if
the +5VSB was only 2A. It's like a weightlifter with
huge biceps, and tiny thin ankles. You decided, to
"make your biceps bigger", by buying a 750W supply.
In fact, the reason you fell over, was your "ankles
were too small". You're looking for a limitation
in the wrong place. The limitation was elsewhere.
If +5VSB is overloaded, and the output drops out,
the entire system shuts off in response.
How do I research these things ? With a clamp-on DC ammeter.
That allows measuring each DC current flow, and determining
current system state.
It is highly unlikely the Geforce 9600 is "drawing 26 amps".
When NVidia quotes a ridiculous number like that, it is
the sum total of:
CPU power (draws from 12V rail) CPU_watts / 12V = CPU amps
Video card (draws mostly from 12V rail) For Skybuck, it is 2.4 amps
Hard drive motor current, about 12V @ 0.6A per drive
Optical drive, about 12V @ 1.0A when media is inserted
Cooling fans, current draw is printed on fan hub, 0.1 amps
You can work out these numbers, and you'll soon see
that your system doesn't even come close to "26 amps".
This is why I recommend calculating the numbers, for
each rail, and working it out for yourself.
I *hate* when people buy power supplies in ignorance.
You can use a power supply sizing web site. At least
one of the sites is getting closer to doing the calculation
properly. The first sites offering this service (doing
the calculation for you), were off by a factor of 2.
I also do calculations on demand for people, when the need
arises. Once you've determined the size, added a
small amount of margin (say 30% margin on 12V rail),
that should be plenty. Even my calculation is
relatively conservative. For example, the boiler
plate value on my optical drive, is 1.5A (2.5A on
Blu Ray). But using my clamp-on ammeter, I checked
one optical drive here, and with media in the
tray it was 1.0 amps. Measurements can add some
refinement to the calculations.
Let's work some CPU numbers, and see how conservative
they are.
I have a 65W CPU in my backup computer system.
The basic calculation is 65W / 12V = 5.42 amps
That's the amount of 12V current, to supply an
estimated 65W TDP for the processor.
The VCore power converter, is a switching power
supply. It is not 100% efficient. To hint that
is the case, when I do power calculations for
people, I pretend the efficiency is 90%. The
actual figure could be 80% or 85%, it's really
hard to guess. So I use a figure like 90%, to
show the conversion is not 100% efficient. Now
my estimate of CPU current becomes
5.42 amps / 0.90 eff = 6.02 amps
Now, let's study the actual processor. I place the
clamp-on ammeter around the two yellow wires of the
ATX12V 2x2 square connector.
65W CPU (E4700 dual core) - should be 6.02 amps
System idle - current flow 1.1 amps
Run Prime95 100% loading - current flow 3.0 amps
So my E4700 only uses 36 watts flat out (12 volts times 3 amps).
Historically, some of the older S478 (Northwood) processors,
they came very close to their TDP value. Some of the
Core2 processors, the lower end ones, were well under
their TDP rating. My E8400, I think it might have
weighed in around 43 watts or so.
Consequently, when I do a power calc, and I use TDP
as an estimate, it may still be over-estimating the
current draw. Neither of my two processors, draws
the TDP value of 65W.
For Skybuck, his system power draw is 200W. There is
no need for a 750W or 1200W supply, they're just
a waste of money.
And don't forget, that all the rails count. Even
if you have a gazillion watt supply, as long as the
motherboard makers insist on running all the USB bus
power from +5VSB, and the power supplies have no more
than 3 amps to offer, there will be instances of
systems shutting off because of it. On the older
motherboards, a series of jumper plugs on the motherboard,
allowed you to run some USB ports from +5V (main rail,
good for perhaps 20 to 30 amps). Instead of the tiny
+5VSB supply. On modern systems, they stripped the
gold header pins, to save $0.10 per system, leaving
a disaster to await, if too much USB loading is present.
*******
A Kill-O-Watt meter is cheap, and allows you to bound
what your PC is doing. If Skybuck owned one of these,
he could see his system power vary from 80W when the
system was idle, to perhaps 200W when gaming. This
would be immediate evidence, that a 750W or 1200W
supply would be a total waste. These are cheap, and
can tell you whether you're even getting close to the
thermal limit of the ATX supply.
http://www.amazon.com/P3-Internatio...F8&qid=1366349216&sr=8-2&keywords=kill-o-watt
When you want to study system power usage, at the
main ATX cables, you use a clamp-on ammeter to get
the amps. Some clamp-on meters, measure AC amps only,
and are used by central air conditioning installers.
My meter, measures both DC amps and AC amps. And the
PC has DC amps on the major DC outputs of the supply.
http://www.extech.com/instruments/resources/manuals/380947_UM.pdf
The benefits of that meter, versus a conventional ammeter.
1) No need to cut any wires to measure current. Simply
place the jaws around the wires and measure.
2) Meter automatically sums the current in a series of
wires. Clamp the jaws around the two yellow wires on
the ATX12V 2x2 CPU power connector. The magnetic fields
around the wires add (as the current flows in the same
direction in each wire). That's how I can measure the CPU
consumption.
3) The meter uses the Hall Effect, to do the measurement.
The meter can measure huge currents, with no thermal
effects. The meter doesn't get hot. Conventional multimeters
are limited, by the shunt resister inside the meter
overheating. That's why a conventional meter has a
fuse inline, which may blow if you overdo it. My clamp-on
meter doesn't have that limitation. For example, I've
measured the starter motor current on my car, at around
120 amps (defective). The meter hardly even notices.
Since the meter has "peak detection", I can catch the
peaks of the starter motor consumption.
I use the 40A DC range, for "inside the PC" work.
I used the 400A DC range, for working on my car.
I use the 40A AC range, when checking my central air.
A more sensitive DC range would be nice on that meter,
but I can live with it.
HTH,
Paul
=D