[Novalug] PSU turns on/off every few seconds

Bryan J Smith b.j.smith@ieee.org
Wed Jan 20 09:08:12 EST 2016


Peter Larsen <peter@peterlarsen.org> wrote:
> So a quick update - new PSU, and with the case still open to allow easy
> access if/when things go bunkers again, "cloud" (the name of the server)
> is still humming along on it's first week. One thing I've learned from
> this is that I way over-dimension my PSUs

Virtually all systems since the early '00s require X amps from +12V
rail(s), as the VRs on the mainboard pull, transform and condition the
rails.  If there is a reduction in current, which _always_ occurs as a
PSU ages (period), voltage will eventually drop to the point there is
not enough potential.

That's why a lot of people don't see issues for months, even years,
and then start running into the case of the board just powering off.
It's when the VR decides to just cut power, instead of resulting in a
brown-out condition and causing temporary, if not permanent, damage.

Plus the VR itself also degrades with time too.  It too, after all, is
it's "own PSU."  Those caps and other components on your mainboard
around your CPU aren't much different from inside the PSU.  ;)

Also since the mid '00s, most CPUs only use the power required, and
turn off units or otherwise run them in low power, "clockless" states
(I won't go into the semiconductor concepts).  So when units get woken
up, suddenly the current can rise, causing the inadequate power
condition, and the voltage drop, etc...

> - I now have an i5 (different micro system) on 80W and it's just humming
> along (well, not now - it's off while I wait for a display cable to be delivered).
> I guess I never paid attention to the wattage required by the mainboard+CPU

Again, the +12V rail(s) is(are) the most important.  Virtually all
mainboards VRs are pulling from it these days.  They don't pull from
the +3.3V or +5V lines.

The +3.3V and +5V lines are largely there for legacy peripherals that
have no on-board VRs.  Most SATA and other drives do though.
Microelectronics for regulating analogy have become very cheap over
the last couple decades.

> - I don't think I have a PSU smaller than 500W on any system.
> Time to reconsider that decision - and just use larger PSUs for systems
> with a lot of drives/perf.

A lot of "aggregate ratings" are crap.  That's why even enthusiast
sites focus on the +12V rail.

Then there's the "efficiency" aspect.  The less efficient a PS, the
more heat it generates.

So even if it claims to output X, from Y input, with added heat, there
is the compounding reduction in output too.  The output is usually
measured at typical room ambient or similar, not when the PS itself
reaches 50C, let alone 90-100C.

> I'm still not convinced the PSU is at fault - I'll have the old one run
> doing some tests later to see what's really going on. Maybe bring it to
> NovaLab since they have osciliscopes etc. needed to figure that one out.

Again, it could be the VR on the mainboard itself degrading.

I probably should have put it that way initially ... your mainboard's
VR is your "other PSU."  Heck, since the early '00s, we've seen a
number of boards (SBCs, not so much PCs, but still PC components)
themselves come with variable 9-27V inputs, so they can take common
+12V or even "street" type +24V, along with notebook A/C adapters
(typically +19-20V, a common voltage largely for charging purposes,
long story).

> Thanks Bryan for your many suggestions and input (as usual).

Not sure I helped, but if I helped you think a little differently,
I'll return a "thank you."

Living in the lightning capital of the world (Orlando), power is
something I deal with constantly.  I got through at least 2-3 UPSes a
year, as I have about 20,000VA aggregate in my home in over a dozen
UPSes.

Although a nearby lightning strike will still fry everything,
regardless of surge and other protections ... including the
underground cabling outside the home.  My parent's nextdoor neighbor
got hit, and everything in a 20 house area was fried, largely through
the cable.

I even had to stop using an APC UPS that was delivering "brown outs"
when it switched to battery power, even after the battery was
replaced.  Pissed me off because it destroyed or damaged another $500
of "replacement" equipment.

I.e., I don't think the lightning strike itself destroyed the original
equipment either, but once they were running on battery, and the UPS
was delivering sub-100V A/C.

When voltage drops, I drop the unit.  Just cheaper to replace than
continue to use.  Brown outs are nasty.  It's also why I hate the
world of Chinese manufacturing, especially with all the UL Listing
forgeries that are dominating the US market (if they have them at
all).


-- 
Bryan J Smith - http://www.linkedin.com/in/bjsmith



More information about the Novalug mailing list