Troubleshooting a possible blown PSU

eltname

eltname

eltname

Associate
Joined
18 Jan 2021
Posts
22
Last night I heard a big pop in the house and all the sockets went off. Had a look around but couldn’t see any obvious issues so unplugged things and switched back the fuse on the RCD. Walked around the house plugging things in and turning things on. Plugged in computer and saw the white light on AIO, the mouse light etc. I pushed the power button on the computer and there was a big flash and a pop of electricity. Fuse tripped again.

Am presuming it’s ‘just’ the PSU blown but what’s a safe way to troubleshoot?

The PSU goes straight to a wall socket. Accessories go to a reliable power bar. That said, when unplugged I noticed it had been plugged into the extension bar in (a possibly expensive) error - would have been for a week at least.

I couldn't smell any burning at the back of the PSU.

The cable itself (a c19) seems fine and the fuse appears fine.

My main concern is preserving integrity of data on the 3 x nvmes.

I’ve ordered a new c19 and a usbc nvme enclosure to arrive today. My plan is to extract the nvmes, check they’re functional on another computer (I assume it’s just plug and play for the two storage nvmes, though not sure what would happen if I attached a system nvme (which has stuff on the desktop) to an existing system?

If they’re functional then at least I can work in the interim while I sort out the PSU.

Computer is less than a year old. I built it and have had no major issues to date. The PSU and many other components were from OCUK.

I presume the PSU will retain some amount of power inside so I don’t want to touch it prematurely.

Parts:

9950x / ProArt x870e / ProArt 4080 x 2 / 96gb 4800 Dominator Titanium (2 x 48gb / T705 4tb / SN850X 8TB x 2 / Thor 1600 Titanium 80+ / ProArt PA602

What am I looking at in terms of next steps?

Take off the side panel (no tiles in sight…), look around for obvious damage, remove nvmes / gpus and try cycling power with new cable in correct wall socket. I have another PSU in a spare machine but it’s not rated for the amount of draw in the current build. Different brand as well so would need to completely recable which would be a blow - I’m self employed and in the middle of a video edit - am fairly sure there’s some recent material which is only on one of the nvmes, so am hopeful they’re not wiped somehow, but they are attached to the mobo which could have failed..

Bugger.
 
Last edited:
urgh that sucks.

first off, remove that psu and either (a) try to RMA it to ocuk (or asus), or (b) bin it if asus don't accept the RMA
in the interim you'd need a new psu at minimum
remove all your ssds and also the gpu

when you get the new psu, disassemble the whole computer and use it like a testbench pc
start with the core components (ie psu, cpu, mobo, ram) with a bios reset on the motherboard...you don't need to connect your gpu at this time as the 9950x has an integrated gpu
if the core components work, then test out the ssds - hopefully they work - then remove the ssds
once the ssds are tested and removed, test out the gpus individually
if all works, thank the god(s) and reassemble the pc
 
Last edited:
Thanks both.

Yep, had planned to replace PSU but just wasn’t sure what steps were in terms of components and if putting existing components (if they WERE fried) on a new PSU would be damaging to new PSU?

Annoyingly the same model PSU is now around 7-800 quid and not easily in stock compared to when I bought it last October for 450 odd!

So I’ll likely get something else - and presumably regardless of what I get I’ll need new cables even if it was a like for like replacement (so another small fortune to cablemod) to power the twin 4080s (if they survived).

I can get one of these delivered tomorrow apparently if nothing better?

Seasonic PRIME TX-1600 Noctua Edition 1600Watt Full Modular 80+ Titanium Power Supply ATX 3.1​


I’ve got all fingers crossed for the nvmes surviving and regardless of whether they do will take it as a lesson to get some ‘ore regular backup processes in place. It’s only about 2 weeks behind but that’s a lot of hours of work potentially lost.

Also maybe my bright idea of having all the fast storage in a single box is now looking like less of a good idea.
 
eltname said:
Yep, had planned to replace PSU but just wasn’t sure what steps were in terms of components and if putting existing components (if they WERE fried) on a new PSU would be damaging to new PSU?
Click to expand...
if the components are fried, they won't damage a working psu
a faulty psu, however is more than capable of damaging working components
steps as i've described in my prior post :)
if any other components have failed, then we'll need to cross the bridge when we get there

eltname said:
Annoyingly the same model PSU is now around 7-800 quid and not easily in stock compared to when I bought it last October for 450 odd!
Click to expand...
if bought from OCUK, RMA the psu to them

eltname said:
I can get one of these delivered tomorrow apparently if nothing better?

Seasonic PRIME TX-1600 Noctua Edition 1600Watt Full Modular 80+ Titanium Power Supply ATX 3.1​

Click to expand...
i wouldn't bother with the noctua edition, that's just extra money for no meaningful benefit, but otherwise that's fine

eltname said:
I’ve got all fingers crossed for the nvmes surviving and regardless of whether they do will take it as a lesson to get some ‘ore regular backup processes in place. It’s only about 2 weeks behind but that’s a lot of hours of work potentially lost.

Also maybe my bright idea of having all the fast storage in a single box is now looking like less of a good idea.
Click to expand...
if mission critical work, always backup backup backup and never to a single box, as you've now learnt, unfortunately the hard way
usb ssd at minimum, or better still, a NAS with RAID
 
Last edited:
tamzzy said:
if mission critical work, always backup backup backup and never to a single box, as you've now learnt, unfortunately the hard way
usb ssd at minimum, or better still, a NAS with RAID
Click to expand...
Yep, it's all backed up and do have RAID but not NAS, as well as cloud backups plus USBC shoot drives (duplicates of each drive etc), footage rushes get backed up to Frame.io at the start of projects, but the project timeline hadn't had a cloud backup in a week and a bit. I have daily exports I can use as a reference to rebuild against, but blimey it's a painful lesson.

Thanks re. Noctua note - though if you've seen my ProArt money for no real reason build, I'm sure you'd realise ******* away money on things that will blow up in the near future is a hobby of mine. I'll take a look on the PSU subforum here and go from there.

Will reach out to OCUK re an RMA
 
eltname said:
but the project timeline hadn't had a cloud backup in a week and a bit
Click to expand...
ouch!

when you said RAID but no NAS i'm assuming you mean the 2x 8tb ssds you have in the main rig? also i assume that the RAID array is set to mirror rather than stripe?
with a NAS you could set up for nightly backups which i'd have thought would be more meaningful as a backup system, just in case the whole system dies (like now, which can take out the ssds)
then the cloud can be the for weekly backups

this way, you have full redundancies in place

normally for mission critical workloads the recommendation is 3 copies of data, which you will have from the above: RAID SSDs in main system (son), RAID NAS (father), cloud (grandfather)
lose one SSD: lose the work in progress
lose both SSDs: lose one days' work (NAS backup)
lose both SSDs and the NAS: lose a week's work
 
Last edited:
No, I've got an external G-RAID studio which, from memory, is RAID 5 with 64TB (so 56TB useable).

The nvmes on board - t705 4tb system/software/downloads (including temporary footage/packaging of projects to hand over which is the stuff that won't be backed up), the 2 x sn850x 8tb are working footage (usually projects are around 4TB, so i have 2 or 3 on the go and space for a few terabytes of resolve cache.

Am sure am missing a trick with automated backups but just never something i've done - has always been manual so on hectic weeks it gets missed. And then computers explode :rolleyes:. And sometimes lessons get learned. Sometimes.
 
ah fair enough then the g-raid serves the same purpose as the NAS - have you checked WD website, they usually have some software available to help with automating backups

eltname said:
Am sure am missing a trick with automated backups but just never something i've done - has always been manual so on hectic weeks it gets missed. And then computers explode :rolleyes:. And sometimes lessons get learned. Sometimes.
Click to expand...
definitely worth some time looking to automate this.
even if it saves your bacon once, it'll have paid back the investment
 
Thanks. Hadn’t heard of these but will try one out. I’ll see if I can get one on Amazon for a morning delivery. Though I’ve just received nvme enclosure so going to try my luck with drives tonight. Then I’ve ordered a new PSU which comes before 10am tomorrow so let’s see, think I’d feel safer with a new one before testing the old. Have asked OCUK about an RMA and have been in touch with insurance in the meantime. This is worse than breaking your phone!
 
NetMan said:
You wasn't using bitlocker / encryption was you?
Click to expand...
No thankfully :)

I presume (if they’re not dead) the 2 x sn850x will just show up as an attached drive on the spare machine. Anyone know what would happen to the system drive? Presume as a system already exists it’ll also just show up as an attached drive but with a windows folder on it etc?
 
eltname said:
No thankfully :)

I presume (if they’re not dead) the 2 x sn850x will just show up as an attached drive on the spare machine. Anyone know what would happen to the system drive? Presume as a system already exists it’ll also just show up as an attached drive but with a windows folder on it etc?
Click to expand...
Boot first then attach it, it will just show as another drive; if you attach it before booting theres a chance it might try to boot from it depending how your boot sequence is configured in the BIOS.
 
NetMan said:
Boot first then attach it, it will just show as another drive; if you attach it before booting theres a chance it might try to boot from it depending how your boot sequence is configured in the BIOS.
Click to expand...
Glad I asked :)

Crossing fingers they show up at all!
 
There's a chance it might not assign a drive letter if that happens you can assign a drive letter using disk management either via computer manager or launching diskmgmt.msc

Newer PSU's are less likely to take out other components when they go bang so hopefully everything is still there and you will just need to RMA your PSU.
 
Last edited:
Thanks all.

PSU is out. It rattles and have shaken out a few little bits of black plastic shrapnel and a tiny thin bit of pink/red film. Both with remains of letters on. Presume some capacitor or something.

I guess my worry re the PSU not protecting the other components is that it obviously blew when it was powered on and then I tried to switch it on again when i was trouble shooting the RCD. But let’s see.

GPUs both taken out. Heat sink is uncovered and off the nvmes and they’re out too. Going to have a go with the enclosure shortly and will do my darnedest not to format the thing in error which is entirely in my wheelhouse.

Other than that the inside looks decent. Certainly not other rattling or obviously singed bits, no burnt smells etc (but there weren’t on the PSU which is obviously knackered so..).

Also, side note, apart from the AIO radiator fins in the top which look a little clogged, I’m really impressed with the dust management the ProArt case / fan pressure delivers.

Side side note - suspect the AIO fins are clogged with dead skin cells from my cat who likes to sit on top - the bonus being that on my previous build he would regularly, I believe intentionally, turn the power button off. The pa602 case has an anti cat lock on the power button which has been invaluable.


Back to reality and a guess my next big fear apart from the nvmes being dead/losing work is wondering what was the cause of the blow out. And was it environmental/something that might happen again (problem with the wall socket for example? It’s oldish - prob at least 10 years if we’ve been in here for 7 and not changed it).
 
They work! Or at least one does, haven’t tried the others yet but first one loaded straight up. Holy **** that’s a relief.

Bodes well for the 4080s and the rest of the bits.

Though genuinely, I would have assumed that two 4080s would be fine on 1600w? The total max draw according to pcpartspicker is something like 1020w.

But regardless, going to get an automated daily cloud backup running tomorrow after a very deep sleep tonight.

 
You must log in or register to reply here.
Back
Top Bottom