Know Your Rails - story of melting death

[Humanoid1]

I've learnt a lot from this whole PSU failing on me thing.

When I built my "current" rig (Q6600 & 8800GT at the time) which has suffered a few upgrades since, I did a certain amount of reading at the time and settled on buying a highly recommended Seasonic brand, and chose a 500watt PSU.
Seems like bags of power to me.

now The poor thing died on me a few days back and seems fairly obvious why to me Now I have read More. Have a look see:

Seasonic S12 500W Silent (SS-500HT Active PFC F3)

+3.3v - 30A
+5v - 30A
(180watt)
+12v1 - 17A - Everything else
+12v2 - 16A - 8/4pin 12v connectors
(396watt)
-12v - 0.8A
(9.6watt)
+5vSB - 2A
(10watt)

Many of you will have noted the rather high power handling 3.3 & 5v rails and pretty weak 12v rails.
Modern systems lean pretty heavily on the 12v rails these days + I had no idea which leads came from which rail so I could even start to consider load balancing them. Combined with my fairly demanding (actually draws the same power while folding etc as the original OC'd 6600 @ 3.51GHz, did a few years with a 5850 @ 5870 speeds folding too) current sig system folding@home 24/7 on CPU and GPU I guess I was asking a little too much!

Some review searching turned up:
12V2 has the 8/4pin +12V connectors, while 12V1 has all the other connectors.
So basically CPU and GPU both on the single 17A 12v (204watt) rail.
At the wall my system was pulling a steady 375watts
- poor thing

I did not realise Seasonic offer as much as a 5 Year warranty (Checking - this PSU had a 3 Year warranty) which I "might" have still been within, or I might have been more hesitant to open up the box. (not recommended unless you know what you are doing as can be dangerous even when these devices have been off for weeks)

I think I know why it "suddenly" died after I have been running this setup on my new rig for 6 or 7 months:

It looks like the 12v regulatory coil heavily overheated (far right coil in far right image on linked page), the insulator on the wrapped copper wire has all turned a nice looking black colour (looks almost like it was made this way, but when new was a bright copper colour) + looks like some of that super hot now soft insulation has caused shorts along its length. Some PSU output leads also melted a bit + where they were touching this coil its insulation is showing through to bare copper on the coils winds.
The nearby (4mm) white plastic fan connector has browned from heat.
A smaller 16v cap is blackened from heat on the side facing the coil
A larger 16v close by cap has was cooked and is mildly swollen (a cap hidden from view in the image I linked).

More likely the slowly cooking caps deviated from spec (perhaps increasing the heat of this coil) outputs leading to the 1st shutdown.
PC restarted just fine when I tried 25 mins later.
Checked event logs + all running voltages etc trying to locate fault - all fine.
Stated GPUz logging to file and started GPU folding
- All fine
Started CPU folding...

all ran fine for maybe 6 mins till next and final death.
(was a little hard to Firmly identify the source of the acrid burning smell when waking in the middle of the night as my room was smelling pretty chemically anyways from working on my motorbike fuel pump - garage is Damn Cold this time of year lol )

Will update with anything interesting from those GPUz logs once the new PSU arrives.

Anyways I just last night ordered an Antec Earthwatts Platinum 650 W Power Supply

+3.3v - 20A
+5v - 18A
(105watt)
+12v1 - 30A - Peripheral, SATA
+12v2 - 30A - EPS
+12v3 - 30A - ATX
+12v4 - 30A - PCIe1, PCIe2
(576watt)
-12v - 0.3A
(3.6watt)
+5vSB - 3A
(15watt)

This looks Much safer and well within specs this time.
This Antec will gain me 10% more efficiency at about 400watt load, a nice little power saving
+ has better volts and ripple too.

I've read of quite a few other people with these old Seasonic S12 500watt PSU's suddenly dying on them + bet they also overloaded their 17A 12v rails.
This is one lesson I won't forget any time soon!

++++++++++++++++++++++++

Kinda hoping I will get a slightly improved OC on my CPU with some luck.
The voltages always were a tiny bit low across the board, think from memory 12v was between 11.86 to 11.89.
I read some folks got a 200MHz OC after updating to better PSU's after suffering a failure.
I had to increase my mobo's PWM frequency to 600k (400k is stock, 800 to 1000k is for extreme LN overclocking I am guessing related circuitry gets hot at higher frequencies so not advised for day to day usage?) to get from 3.6GHz to 3.8GHz.
3.6GHz was really easy at stock voltages etc.
Getting beyond 3.6GHz and remaining prime 95 + folding stable was Really Hard even with high CPU volts (testing about ALL possible options and combinations - learnt a LOT about this chipset + how it all inter-relates with its various sections and CPU areas lol).
Was a weird OCing experience, with unreliable higher than 3.6GHz attempts. Sometimes 4GHz seemed stable at stock volts, other times at High volts it was Hard/impossible to boot let alone make stable.
Increasing PWM frequency was the Only thing that worked reliably.
I am feeling now that the struggling PSU was responsible for weird reliability at over 3.6GHz.

Hoping for good things from this upgrade!

Sig PC at time of writing:
Antec P182
Gigabyte X58A-OC
Xeon X5650 @ 3.8GHz 1.23v(idle30c folding62c) / Noctua NH-D14 (140mm fan only @ 1.1k rpm)
3x 4GB Kingston HyperX LoVo 1,600MHz 8-9-8-24 1T @ 1.5v
Asus 7950 Direct CUII TOP @ 1.1GHz 1.1v
Creative SB X-FI Xtreme Gamer Fatal1ty Pro
4x HDD's ~ 3.8 TB worth
Seasonic S12 500W Silent

 

[Bobnova]

Definitely a good idea to check out where the rails go and such!
The 4/8pin 12 V connector powers the CPU, that PSU was to the Intel spec at the time which called for the CPU to get its own rail an everything else to share the second rail. It's the spec that kicked off the whole "singe rail is better!" thing. Intel revised the spec.

 

[irrykuxeh]

Ya, if it's too much work to arrange components on to rails that will supply enough power, you can always buy a PSU with a single 12V rail rated for some ludicrous amperage.

Otherwise, a little research may be required to avoid melting death

 

[Humanoid1]

I see, So that would equate to the Seasonic S12 500w having:
(actually pretty hard to find out. Only EXTREMEoc mentioned it that I saw )

+12v1 - 17A
1 x 20+4pin main ATX power cable
2 x 6pin PCI-Express connectors (single cable)
4 x Serial ATA connectors (2 cables)
6 x 4pin (Molex) connectors (2 cables)
2 x floppy connectors (2 cables)

+12v2 - 16A
1 x 4pin +12Vconnector
1 x 8pin +12V EPS connector

So a little strange that it died if the CPU was primarily powered by the EPS connector on the 2nd rail.
But I think also a big part of the problem was that dust (even though my P182 case has regularly cleaned dust filters) had actually collected a fair bit around the PSU output cables right over that coil - must have had Very Little airflow there
(I had it in my mind that the CPU was mainly powered from the ATX cable, due to a very missleading and contradictory line from the Gigabyte manual that implied that the EPS was not even needed + was only recommended for Intel Extreme Edition CPUs (I had it plugged in anyways ofc))

Most cases these days still often don't have dust filters. Must cause havoc for those PSU's, especially when run 24/7 over the years.


@Atomichearse
Indeed so, at the time I had not gone much beyond good brands + overall wattage systems needed in reality. Just knew that 500watt was Loads for my system... just had missed the Amps per rail putting overwhelming faith in the Seasonic name to make a PSU that would easily do what was needed.
I suspect if that dust had not built up in there, it would have been fine today still.
- I did fairly frequently look in through the rear grill of the PSU, but it looked pretty clean. Only could see that corner of dust once I disassembled the PSU.

 

[Bobnova]

The "rails" are just individual OCP anyway, all 12V is generated in one place and then is split for OCP reasons. Odds are dust and a loooooong heavy load is what did it in.

 

[irrykuxeh]

@Atomichearse
Indeed so, at the time I had not gone much beyond good brands + overall wattage systems needed in reality. Just new that 500watt was Loads for my system... just had missed the Amps per rail putting overwhelming faith in the Seasonic name to make a PSU that would easily do what was needed.
I suspect if that dust had not built up in there, it would have been fine today still.

Yeah, I did the same. I put all of my faith in Seasonic knowing that it was a good brand, a good efficiency, and that 750W was more than I needed. Long story short: It died after about 3 months anyway. And it was a single 62A rail. I posted about it here not too long ago in Gen. Hardware. Sometimes you just get unlucky :/

 

[Humanoid1]

Yup, I think that's it.
I did not immediately finger the dust as the cause as there is relatively little of it. Thinking back over it (was hoovered up as 1st action after opening up PSU) the little dust there was, unfortunately strategically blocked almost All airflow to the area around the overheating coil.

Its extremely blocked in at the best of times by wires, side of PSU case + heatsink overt he top + around it. Unfortunate design that really could have had a hole in a small area of that heatsink to stop this happening.

I guess the only way to clean the damn things out without voiding warranty is to blow the dust out with a compressed air line. Or perhaps air in a can that has a long nozzle to direct it inside to hard to reach areas. Especially for folks that do not have case dust filters!