Bill Anderson wrote:
The saga continues:
When I built a new computer around an Asus P5Q Pro Turbo back in 2007, I
acquired a PC Power and Cooling Silencer 610 EPS12V which I assumed
would meet my needs forever because as we all know, computer components
never get outdated. And then when I built a new system around an Asus
P9X79, I found my good old PS continued to meet my needs. And now I've
added a Kraken X61 to the mix, along with a Silverstone USB 3.0 internal
card, and at first all seemed well. But then I decided that with all
this extra cooling capacity, maybe it was time to see what my P9X79
could do if I overclocked it. So then I installed the Asus AI Suite II
and told it to overclock things as it best saw fit. And then I started
getting occasional voltage warnings in the lower right corner of my
screen. Seldom the same warnings -- sometimes voltage 5.0 something,
sometimes 3.something ... like this:
https://www.amazon.com/clouddrive/sh...hare_link_copy
https://goo.gl/iGE3IK
Well that's no good, so I told AI Suite II to stop doing what it was
doing, and after a few scary reboots in which the system told me
overclocking had failed and I needed to go into BIOS and do something
without explaining exactly what I should do, things seemed to get back
to normal. But no...still, occasionally, I get a warning like the one
above, even though I think I've reset BIOS to the way it was before AI
II and the Kraken installation.
Every suggestion I've Googled ends up saying my PS is no good. My
wonderful PC Power and Cooling power supply from 2007? That power
supply is no good? Just because I've recently hung some new fans, a
pump, and a USB 3.0 card off it? My faithful PS has finally met its
match? Really? I actually need a power supply better than the one I
have? You think?
It's a strange design, in that it is an early 80+ product,
yet it has a generous 3.3V and 5V rail on it.
80 PLUS Certified Active PFC
http://images10.newegg.com/NeweggIma...703-005-04.jpg
Current generation 80+ products, use two stage conversion. The main
supply is 12V. Then, it is followed by a small 12VDC to 3.3V/5V board.
And those typically give 3.3V at 20A and 5V at 20A. And the combined
rating is usually barely sufficient to allow one of those two
rails to be fully loaded (100W to 150W combined).
The low rails on yours is rated 170W combined. And the spec plate
makes it look like a single output supply. (Even though, for
safety reasons, there may be a 20A limit on subsections of the
12V loom. Maybe 12V1 and 12V2 have 20amp limiters, and whatever
it has for PCI Express 2x3 or 2x4 would be on a separate
limiter. It isn't a good idea to let loose a 50A max
current flow, on spindly 20 gauge wire.
From a spec perspective, the supply doesn't hint at being
a lightweight.
*******
There was a time, when you couldn't trust Asus Probe.
But the design of these things has changed a bit. At some
point, the scale factors for the measurement circuits was
put into a BIOS table. Which should reduce the coding errors
when it comes to scaling readings from the hardware monitor.
You could always use a multimeter, to check the rail voltages
and see how much they're in error.
The 3.3V should be accessible on a SATA power connector. That's
if you don't want to go after the ATX main connector. I would drop
into the computer store, get an extender cable for SATA power of
some sort, cut the 3.3V wire, and make a measurement there. Being
careful to not leave that exposed wire out where it can do harm.
I think the SATA on my supply, has the full five wire harness
and so 3.3V is on there. When you use Molex to SATA converters,
one thing you don't get that way, is any 3.3V signal (as that's not
on the Molex).
Could the pump be too much for the thing ? Maybe. But if
Asus Probe is sounding those errors when the machine isn't
at 100% load, then you'd better give it a thorough examination.
In terms of test gear now, I have my voltmeter ($20 type),
I have a clamp-on DC ammeter ($300 when I got it), and the
Kill-a-Watt meter (maybe $40CDN when I got it last month).
The clamp-on ammeter is particularly nice, in that as long
as the main power bundle is physically accessible, you can check
current flow on each rail (place all four wires of one color, into
the jaws of the meter at the same time). The meter can sum the
magnetic field around all four wires, and give you a total DC amps
for that rail. You can then go off and check the SATA or IDE Molex
cable current flows, total it all up, and see if it exceeds the spec
rating of the supply.
I didn't do that on my last build. When I burned myself on the
VCore regulator, I did get out the clamp-on meter to check the
ATX12V current flow (12V @ 13A). But I was more focused on getting
the VCore cooling there in good shape, than doing a complete
characterization.
Since you're getting alarms on 3.3V and 5V, that suggests
a two stage regulator, and a small DC board for the low rails.
Yet the spec rating is so generous, the specs suggests one
large transformer runs everything, and the voltages are
kept in step via the turns ratio of the transformer. If
anything, on a unified ATX, if the 12V was feeling the strain,
the feedback would turn up the primary, and all three rails would
rise. Since the 12V is loaded, it comes in just at spec, while
the 3.3V and 5V in such a case, would be on the high side.
This is called "cross load regulation", the ability to stay within
the ATX spec limits, when ever there is a skewed load on the
main outputs. On supplies where actual separate circuits are
used for each rail (just a few of those were made), the cross load
is down around 1%. Because without a unified transformer, there is
no cross loading to speak of.
And on the two regulator designs, there really shouldn't be a
classic cross load behavior to speak of either. The 3.3V and 5V
could affect one another (load 5V, 3.3V pops up). But the 12V and
3.3V/5V board, shouldn't correlate in terms of "voltage shift
direction".
*******
Just get your multimeter out, and check it :-)
I wouldn't ignore the warning entirely, especially if
it's something easy to fix.
*******
It's a bit difficult to get specs for Kraken x61.
I tried the manufacturer web site but couldn't find
what I wanted. This review article picture, shows at
least one spec is on the side of the box. The pump is
12V @ 0.325A. The fans don't have specs, but I'm going
to pretend the two fans draw 0.3A each. So the whole
kit is around 12V @ 1A. That should be like a flea
sitting on your power rails. The only bad thing about
motors, is back EMF into the rails (load ripple), which
might be a problem if the supply was really lightly
loaded.
http://media.bestofmicro.com/C/T/438...l/DSCN0340.JPG
I'm going to have to blame a phase of the moon for
this one... I don't see a reason for a change in behavior.
Paul