So, my PSU just melted

iviv

iviv

iviv

Soldato
Joined
2 Jun 2007
Posts
6,839
Location
Mornington Crescent
Specs: Seasonic focus plus gold 1000w psu
Powercolor red devil vega 64
Amd 3900x
Gigabyte X570 Aorus Pro

I had been dealing with some stability issues for the last couple of months and had been trying to rule various components out, switching out almsot everything while struggling to pinpoint the actual issue, but never actually touching the PSU since it never occured to me that it could be the problem. Long story short, yesterday I suddenly experienced a ton of graphical corruption and then the PC switched off, and I could smell a burning plastic smell. Looked into my pc case and (images spoilered due to size):
IMG_20200517_202414_1.jpg

SPOILER_IMG_20200517_214835.jpg

SPOILER_IMG_20200517_214644.jpg

The cable from PSU to GPU was hot to the touch, had melted off the insulation at the PSU end of the cable, and had also melted some pins from the connector, requiring some serious tugging to separate.
Thankfully I had recently ordered a psu for a NAS and I've been able to swap that in to keep my main rig going in the meantime.

Doing some googling, I found this article from guru3d which notes that units with serial numbers from before Jan 18 could shutdown under heavy load. This matches up with what had been happening over the last few months, though I've not been able to find a way to tell the production date from the serial number.

OcUK directed me to Seasonic for RMA due to it being purchased over a year ago (Aug 18), and Seasonic were quick to approve the RMA and send a shipping label, but I noticed that I have to cover the cost of shipping it back to them.

So, I've got a couple of followup questions from what happened. Firstly, is it normal for the customer to pay the shipping cost to them? I wouldn't mind so much if it was the case of paying to ship it to them, and Seasonic then refunding the shipping cost if the unit was indeed faulty, but when a power supply literally melts, I feel a little miffed that I'm ending up out of pocket as a result. I've not prepared a box to ship it in yet, since I don't have the original packing to hand, but given the weight, I'm imagining it won't be too cheap to ship.

Secondly, I'm a little concerned about the rest of my system. From some testing and benchmarking with this backup psu, everything does appear fine. The main component I would be worried about is the graphics card since it was the GPU cable that melted. As mentioned, it appears to benchmark ok, so I'm assuming nothing is wrong, but I'm a little concerned that down the line if it starts playing up, could it have been caused by the PSU dying like this? Or would the various protections built in to it have caught any danger and limited the damage just to the PSU itself?

Thanks for your time and happy to answer any followup questions.
 
Were you powering the gpu from a single lead, as in both daisy chained connectors on a single lead? If so then that is most likely what has caused this. Vega and especially Vega 64 has stability problems when powered by a single lead from previous threads I have seen on here and all were cured by powering the card with two seperate leads. What it looks like has happened is that due to being powered by a single lead the gpu has tried pulling too much power for the lead which led to it heating up and melting the insulation and plug. Always use two seperate leads when powering a power hungry card.

Yes, outside the first year the sender is responsible for postage costs and it's quite normal. Some companies such as Gigabyte charge you for the return postage. Inside the first year we should always go through the place it was purchased from and should get postage fee's refunded.
 
pastymuncher said:
Were you powering the gpu from a single lead, as in both daisy chained connectors on a single lead? If so then that is most likely what has caused this. Vega and especially Vega 64 has stability problems when powered by a single lead from previous threads I have seen on here and all were cured by powering the card with two seperate leads. What it looks like has happened is that due to being powered by a single lead the gpu has tried pulling too much power for the lead which led to it heating up and melting the insulation and plug. Always use two seperate leads when powering a power hungry card.

Yes, outside the first year the sender is responsible for postage costs and it's quite normal. Some companies such as Gigabyte charge you for the return postage. Inside the first year we should always go through the place it was purchased from and should get postage fee's refunded.
Click to expand...

Yeah, I'd come to that conclusion myself. For cable management I'd just gone with the single cable on the assumption that a) the cable would be able to meet the ATX spec on both 8-pins at the same time and b) the gpu wouldn't pull more than the ATX spec over each 8-pin. Not sure which of those assumptions was incorrect (Or why it took 18 months for this to happen), but I'll definitely be using separate cables with the replacement psu.
 
iviv said:
So, I've got a couple of followup questions from what happened. Firstly, is it normal for the customer to pay the shipping cost to them?
Click to expand...

Yes, just as you would pay transport costs if you were taking it to a store. But I have in the back of my mind that for defective goods like this you must be reimbursed reasonable expenses for postage & travel. Someone with knowledge of the Sale of Goods Act & similar will undoubtedly chime in shortly.

iviv said:
Yeah, I'd come to that conclusion myself. For cable management I'd just gone with the single cable on the assumption that a) the cable would be able to meet the ATX spec on both 8-pins at the same time
Click to expand...

Given that this PSU is designed for multiple GPUs - it's a 1000W unit - it should be able to cope with 2x 150W on one cable, so failing to do so is definitely a fault. Just imagine if it were running two different GPU-intensive tasks in two VMs.
 
Quartz said:
Yes, just as you would pay transport costs if you were taking it to a store. But I have in the back of my mind that for defective goods like this you must be reimbursed reasonable expenses for postage & travel. Someone with knowledge of the Sale of Goods Act & similar will undoubtedly chime in shortly.



Given that this PSU is designed for multiple GPUs - it's a 1000W unit - it should be able to cope with 2x 150W on one cable, so failing to do so is definitely a fault. Just imagine if it were running two different GPU-intensive tasks in two VMs.
Click to expand...

I didn't reply earlier but I had the same thoughts.

A) It stands to reason that if a high power PSU has 2x PCI power connectors on 1 cable, the plug and cable gauge should be able to cope...
B) It also stands to reason that if you have a GPU with 2 PCI power sockets you should use 2 separate physical cables where possible to spread the load, but that's not really specified from what I can see, the fact you can have 2 full on GPU connectors on one cable is worrying unless the cable gauge is pretty thick, and that could be a crucial issue as demonstrated in the pic above.
C) I wonder if graphics cards at the meatier end of the market exceed ATX spec, or if it's even specified in ATX spec? without downloading ATX spec documents, I'd hazard cable gauges and plug are not.

or
D) the cable/plug could simply be defective/freak accident.


Either way, your lucky that didn't turn into a house fire!!
 
Quartz said:
Yes, just as you would pay transport costs if you were taking it to a store. But I have in the back of my mind that for defective goods like this you must be reimbursed reasonable expenses for postage & travel. Someone with knowledge of the Sale of Goods Act & similar will undoubtedly chime in shortly.
Click to expand...

I'm pretty sure the sale of goods act only covers you within 28 days, everything outside of that would be returned at the senders expense. The company I work for operate in a similar way however if the failure is proven to be due to component failure rather than user error, that return carriage costs are refunded, but that's my company, others might vary in their policy.
 
Dave Burton said:
I'm pretty sure the sale of goods act only covers you within 28 days, everything outside of that would be returned at the senders expense. The company I work for operate in a similar way however if the failure is proven to be due to component failure rather than user error, that return carriage costs are refunded, but that's my company, others might vary in their policy.
Click to expand...

It's 6 months, and it's the consumer rights act 2015, unless purchased pre-2015, in which case the older sale of goods act would still apply.

Company policy is irrelevant, it's a statutory right.
 
mattyfez said:
I didn't reply earlier but I had the same thoughts.

A) It stands to reason that if a high power PSU has 2x PCI power connectors on 1 cable, the plug and cable gauge should be able to cope...
B) It also stands to reason that if you have a GPU with 2 PCI power sockets you should use 2 separate physical cables where possible to spread the load, but that's not really specified from what I can see, the fact you can have 2 full on GPU connectors on one cable is worrying unless the cable gauge is pretty thick, and that could be a crucial issue as demonstrated in the pic above.
C) I wonder if graphics cards at the meatier end of the market exceed ATX spec, or if it's even specified in ATX spec? without downloading ATX spec documents, I'd hazard cable gauges and plug are not.

or
D) the cable/plug could simply be defective/freak accident.


Either way, your lucky that didn't turn into a house fire!!
Click to expand...

Yeah, like, if it was an 8 and a 6 pin on one cable then sure, I would use two of the cables and the 8 pin from each, rather than doing something stupid with the 6 pin and an adapter to turn it into an 8 pin. If you're putting two 8 pins on the same cable, using a reputable branded high power PSU, then I would expect that it should be able to power them.

THW's review of the gpu shows an absolute stock peak of 364W power draw, a smidge under the combined 2x8 pin + pci-e slot 375W supply. The peak OC mode apparently hit 400W, though I was running the card stock without any extra OC or power boosting.

Seconding the comments about being glad I'm not on fire though. Started looking around my flat and realising I'm not entirely sure what I would have done to try and put it out. Maybe just unplug it at the wall, pull cables from the back and drag the case out to the fire escape as long as the flames weren't too bad. Maybe I should grab a small fire extinguisher or blanket.
 
Looks over at Radeon 7 plugged into seasonic supply with a single cable :cool: It's just so much more aesthetically pleasing with 1 cable.

To be fair I did do exactly this to a kolink supply on one of my mining rigs. The smell is special :)
 
iviv said:
Yeah, like, if it was an 8 and a 6 pin on one cable then sure, I would use two of the cables and the 8 pin from each, rather than doing something stupid with the 6 pin and an adapter to turn it into an 8 pin. If you're putting two 8 pins on the same cable, using a reputable branded high power PSU, then I would expect that it should be able to power them.
Click to expand...

Yeh that's kind what I was saying in a round about way, my new PSU came with 3x PCI cables, and it made me raise my eyebrows because they are not 3 cables with 6+2, each one has Two X 6+2 connectors on the end of it...
 
I just picked up a Seasonic GM-500 for the eldests PC, only has one PCI-E outlet with 8 pin to twin 8 pin. Only an RX580 so no issues on overall load.

The bigger surprise is the GM-550 and GM-650 are exactly the same with only one PCI outlet.

Even the +12v ATX cables are a little on the thin side compared to Corsair.
 
@decto At the end of the day it is only a budget psu which is most likely why they have cut back on connectors. Over at Toms hardware there is a large thread on psu's in their forums and one of the guys sent a email off to Seasonic questioning them about the placement of the core series in their lineup and got this reply:-

The main purpose of the CORE series is for system builders where fewer connectors and lesser performance are needed and the price point more important.
Click to expand...

It appears that they had the same idea as Corsair did years ago with their "builders series" psu's. What I find strange is that I cannot find a review on any Seasonic Core series unit anywhere.
 
Quartz said:
Given that this PSU is designed for multiple GPUs - it's a 1000W unit - it should be able to cope with 2x 150W on one cable, so failing to do so is definitely a fault. Just imagine if it were running two different GPU-intensive tasks in two VMs.
Click to expand...
iviv said:
THW's review of the gpu shows an absolute stock peak of 364W power draw, a smidge under the combined 2x8 pin + pci-e slot 375W supply. The peak OC mode apparently hit 400W, though I was running the card stock without any extra OC or power boosting.
Click to expand...

THW also shows that card only uses 30W from the PCI-E slot (which is a good thing, you don't want your mobo overloading), so the remaining power is coming from your PCI-E Cables, so it could easily be going over the 300W spec, with spikes over 350W. If that's sustained over 300W, plus spikes it's not that surprising it melted.

Another factor is the ambient case temps could have been the tipping point. Do you know what case temps have been like recently and were they higher than in the winter months?

It's a bad idea to run any part of the PSU to it's limits and in this case probably over them.

The GPU could also be performing worse than THW sample, over time power consumption could increase, probably only 1-2%, but as your already on the limit even such a small amount could be enough.
 
seasonic are a top quality power supply but i had a similar unit just blow up in less then a year, thankfully rest of pc was still fine, threw an old corsair ps in and it was fine. Think its unfair you have to pay cost to send it back- a refund after they diagnose it is their unit and not misuse would be fair but it shouldent be on you for their fault.
 
Looks to me like the electrical load through the pin was just in excess of what the pin was rated at.
As it got hot, it would have tarnished and increased the resistance.
As the resistance went up, it would have got hotter, and so on etc etc./
Eventually, it would have so hot that it melted/burned.

I would suggest you don't tell them that you daisy-chained the cable because they might reject your RMA.
 
You must log in or register to reply here.
Back
Top Bottom