How/why would a PSU would catch fire/short whilst PC is powered off?

[JLV]

Hello all!

So, as the title says, had a rather interesting experience a few days ago.

Had some time off last week, came back from hols after being away about 4 days or so, as is normal I kill the power to most things in the house whilst away(I assume this is normal!), anyway, Friday AM I spend a bit of time on the pc, nothing intensive at all, playing a few games and watching some netflix after, nothing too exciting.

Wife and I decide to go do weekly shop, so I power down PC and only a couple of minutes later whilst I am getting ready (maybe 5-10 mins absolute max), my wife yells through from the hallway that she can smell a metal burning smell, as I made my way through, I could also smell it (kind of like a soldering type burnt electronics smell). Not knowing where it was coming from, I immediately went through to our main circuit breaker in the hallway (as this was closest to me at the time) - checked it and it looked and smelled just fine so to speak. As soon as I went into our study room (where my gaming PC is located and this was the next closest room) I instantly realised the smell was getting stronger and must be coming from there. I quickly checked the peripherals such as the monitors and other plugged-in electronics before noticing that my actual PC case was almost glowing slightly blueish white near the bottom of the case (as there is a see-through side panel on the PC case).

As soon as I opened the side panel to investigate, I could smell and see the issue was originating from the power supply unit within the PC, worryingly I could also at this stage hear some faint crackling (assumedly as the capacitors were popping and burning away within the PSU) and what can only be described as hearing a strong "electrical current" like something out of a movie about Tesla coils so to speak. I instantly removed the power cable from the PC and moved the entire unit away from other valuables in the room as there are multiple monitors, sound bars etc.

Anyway, with that background out of the way, the question I have now is what on earth could have caused this?

The PC was completely shut down, sits elevated on my desk (still had power connected ofc) but the only things drawing power at the time would have been the actual Mobo rgb lights that stay on when the PC is off and I was charging a Logitech wireless headset via a USB slot on the Mobo itself. The PSU was plugged into a surge protector (admittedly this surge protector is quite old, probably about 5-6 years old) Specs for PC below:

CPU: Ryzen 5900x, 1 year old,
GPU: ASUS Rog STRIX AMD 6800XT LC OC 16Gb, 1 year old
MOBO: Asus Rog Crosshair VIII HERO Wi-Fi X570, 1.5 years old
NVE Drives: 2x Samsung 970 EVO Plus 1TB drives, 1 to 1.5 years old
SATA Drives: 2x Samsung 1TB Evo SSD, 3 years old
HDD’s – 2x Seagate barracuda 10000rpm drives, probably 5-6 years old
AIO Cooler: Corsair Hydro H150i PRO RGB 360mm, 3.5 – 4 years old
RAM: G.Skill Trident Z RGB 32GB (4x8GB) DDR4 3200MHz Quad channel kit, 3.5 years old
Wireless card: Asus PCE-AC88 Dual Band AC3100 PCIe adapter, 3.5 years old
PSU: Corsair HX1000I 1000W PSU, initially bought 7 years ago, however had a replacement one under warranty sent out in December 2020
Case: Corsair 900D Case, 7 years old

Unfortunately, the majority of the above listed parts are now "written off" so to speak, have a few spare mobo's and a PSU from previous builds so was able to test the main components and the casualty list is pretty much everything - aside from the Wireless card, 1 of the 4 Ram sticks and the case. Also, before anyone asks or perhaps assuming some nasty overclock business, there was no overclock configured to this system, never has been and has been running stock since I put it together after upgrading the CPU, Mobo and GPU last year.

Really hoping someone can help me understand what possibly could have happened?

 

[JLV]

Both fair replies:

Finners - I had no idea about Windows (currently on W11) only really hibernating so to speak (when I refer to the RGB I literally mean the RGB on the Hero board itself - nothing expansive on my side but take the point that power delivery was clearly still there!)
gpuerilla - fair point however will say that the PC gets cleaned internally about every month (air cleaner - not compressed air and all filters get pulled and cleaned)

My only real concern(aside from the fact that its gonna be nice and costly to replace the entire build) is had we actually left the house as planned, what could have happened, as no joke or over reaction here, but had I not found it and yanked power when I did it was already smoking and crackling so assuming next step was full on flames and not just the tiny little sparks that were going when I arrived! Was very surprised that both the PSU and more concerningly my electrical mains breaker and surge protector didn't cut power but then again I am not an electrician so only assuming these things should have some sort of voltage/current protection or cut off to prevent this type of issue.

I have out of Curiosity plugged the PSU back in (of course not within the build or house but outside on an extension to see what happens) and it immediately starts smoking again and starts that electrical current noise, this is without anything else attached or drawing from it so I do think there is some sort of regulation issue within the unit itself that must have failed and not necessarily something else within my PC that failed and drew too much etc if that makes sense? although it may be that any potential regulation it had could now be gone so may still have been another cause (not that it matters as its all fried now anyway!)

 

[JLV]

I have seen a few PSU failures in my time from both ATX and the HP server types. One of them was under full load and I was present so isolated and opened up the windows and let the smoke out limiting the damage to other stuff. The other times the PSU generally pops and fizzles but apart from that horrible smoke has not caused any further damage such as a fire. Most when they fail (for me) was when starting up the system.

This does make you think however placing the tower in such as place that it wont spread to other stuff is a good idea. Like you say if nobody was present how bad could it have gotten?

Exactly - would not have been great had I not been home/noticed the smell before leaving - really unsettling that it happened without any load (well standby/hibernating load only) - guess I will now pull power cable every time I turn off a pc in future

Same, have had a couple PSU's fail before in my personal builds, like 2 total to be fair across the last 20 years or so, but as you say, usually dies and then won't boot/start-up but usually problems are limited to just the PSU and maybe mobo at worst in my previous experiences - hence why I was hoping that most of the components would be ok, but unfortunately mainly the GPU and CPU are both dead after having tested them on a different build being fortunate enough to still have my previous build kicking around as a personal gaming server rig.

 

[JLV]

Corsair are usually pretty good with warranty stuff, if you let them know I'd imagine they'd like to investigate.

Yeah - they were good with my previous PSU that failed, will send this one back too but doubt they cover the rest of the system with the limited warranty I assume.

 

[JLV]

I think they have done in the past, so I'd tell them about it, but obviously no guarantees.

fair enough, worth a try thanks!

 

[JLV]

ATX PSUs have 5V standby always powered when receiving mains power.
That's what powers mobo's control logic detecting your power button press and then pulls PS_ON signal wire of PSU to ground for full start of PSU.
Also USBs and any LEDs normally powered by main 5V are powered by 5Vsb when PC is shut down.

Anything else is completely shut down and even during operation there's galvanic isolation.
Which is designed to be good enough to block 3kV short time and 1.5kV in general into mains input from reaching low voltage outputs.
While actual 5Vsb source should have protections monitoring output voltage.

And while mobo could be killed by 5Vsb, graphics card is connected only to 12V and through PCIe slot to 3.3V.
So death of graphics card hints to larger failure than 5Vsb.
Possibly something shorted mains input side to low voltage output side.
Anyway crackling sounds hint to high voltage/mains short.


Would be interesting to see insides, but Corsair should be interested on looking what happened.

Thanks, interesting read and certainly gives me a bit more info to consider.

Not sure if this would provide more context, but whilst this was all happening (i.e. before I yanked the mains cable out the back) I noticed the mobo rgb lights were still on and went off rather quickly after I disconnected mains - I also immediately removed the GPU power cables (2 Independent 8 pin connectors) and noticed a red light on the actual GPU right next to the power input go on for a split second before all power was gone within the system - so I am assuming (perhaps wrongly) that whatever input failure happened wasn't stopped by any of the PSU's safety measures to protect or isolate the system from any sort of voltage spike etc - as given you point out that the GPU would run off of an entirely different voltage to that of the standby 5Vsb

Not sure if that thought process of mine is even relevant but just trying to figure out how the GPU died also (given my previous experience with failed PSU's have always been during use and at worst resulted in a dead motherboard in one of the two instances - this is the first almost system wide failure of so many components at once that I have experienced or heard of really)

All the same, aside from getting in touch with Corsair to discuss an RMA for the PSU at the very least - and hoping to try my luck with some of the other components that died, I have also arranged for an electrician to inspect our mains breaker etc to ensure there isn't a bigger issue with wiring perhaps etc (unlikely but better safe than sorry I guess!)

 

[JLV]

could have been starting as you powered down. then carried on burning. i once had this while playing bf many years ago. could smell like you did and then finally looked at case and full on flames were coming from the psu. it still was in bf lol. so turned it off quickly. then took out psu. was a ocz one. luckily nothing else got damaged.

fair point, I have considered this as a possible explanation as this all happened within 5-10 mins max of me powering down so a very likely scenario, at which point what EsaT said regarding the standby power provision could have been enough to keep things going south once it was powered down... my only frustration is that it isn't a very "safe" failure to have from a reputable PSU supplier (or at least in my experience Corsair have been decent) if anything I would hope this is the root cause rather than a wider problem that is yet undetected within the house electricals etc. (sounds paranoid I know but in all honesty not a nice experience, still keep thinking how bad it could have been had we not caught it when we did!)

 

[JLV]

I believe that Corsair may replace your components if it is found that the psu was at fault. It has happened in the past and is looked at on a case by case basis so it's worth mentioning that you have lost everything due to a psu fire especailly as that lot is worth a quite substantial sum of money.

Yeah thanks, certainly have mentioned it in my RMA request to Corsair over the weekend, did a calculation on the component cost (in the event I go via my contents insurance policy) and if I literally replaced the entire build including the case it would be just about £4k (prices taken from time of purchase so some are cheaper, i.e. I bought my RAM at a terrible time!, but mainly some are a bit more expensive or can't be purchased anymore such as the GPU is now £400 odd more than it was when I bought it)

Certainly worth a shout asking Corsair nicely

Will update here for any future queries on if PSU's supplier take liability and secondly if not how easy it is via home contents insurance (which I am not expecting it to be!)

 

[JLV]

Graphics card might have indicator for missing 12V connector.
And in this case it's for that:

"Piirilevy on varustettu kahdella 8-pinnisellä PCI Express -lisävirtaliittimellä ja niiden kohdalla piirilevyllä on ledi ilmaisemassa punaisena, jos virtaliitintä ei ole kytketty paikoilleen."​

https://www.io-tech.fi/artikkelit/testissa-asus-rog-strix-lc-rx-6800-xt-oc/

(Google's translation would be actually for once very precisely correct)

So card was getting some power to be able to run circuitry monitoring presence of external 12V connectors connectors.
Not sure from where that came, but might be mobo using 5Vsb to power some VRM giving basic 3.3V for slots.
(Wake on function with PCI-e NIC would need power in stand by)

Maybe that was the route how PSU' killed graphics card.
In sleep mode 5Vsb would be used also for powering RAM and that connection could have let overvoltage in 5Vsb to kill DIMMs.

Thanks for the extra explanation EsaT - will post back if/when corsair come back to me with an update on the PSU investigation.