PC just caught fire

[TheRandomGuy7]

I just realized there could be another issue here.

While some high end PSUs have plenty of individual PCI-E outputs, they are often paired into rails.

My EVGA has 3 rows of 2 PCI-E outputs, it could be that even if using two cables, they need to go onto different rows.

Looking at a pic of the outputs on OP's PSU, it just has 6 PCI-E and CPU outputs all in a block, it could be that the outputs are shared in the same manner that a daisy chain splitter would be?

The problem with daisy chaining cables isn't that the PSU gets overloaded on the same rail or anything. It's that the cables themselves are only rated for 150W or so so a lot more than that can cause them to melt/catch fire.

You did get me thinking for a second because my GPU is definitely plugged into 2 PCI slots on the same 'row.' But I really don't think there's an issue here and it would be ridiculous oversight for the PSU manufacturers to not even mention it.

 

[BallistixOnZ490]

The problem with daisy chaining cables isn't that the PSU gets overloaded on the same rail or anything. It's that the cables themselves are only rated for 150W or so so a lot more than that can cause them to melt/catch fire.

You did get me thinking for a second because my GPU is definitely plugged into 2 PCI slots on the same 'row.' But I really don't think there's an issue here and it would be ridiculous oversight for the PSU manufacturers to not even mention it.

Except it wouldn't be mentioned for older PSUs that were not built to the power delivery requirements of the newest cards.

Its still something that people should be overly cautious about regardless.

 

[TheRandomGuy7]

Except it wouldn't be mentioned for older PSUs that were not built to the power delivery requirements of the newest cards.

Its still something that people should be overly cautious about regardless.

I think OP mentioned it was an RM1000x though so not really old nor unfit for high power GPUs.

Agree that no one should take any risks with PSUs, and that includes only using high quality units and never daisy chaining cables. I tried googling around to find info on rails but there really isn't any info out there, I'm hoping you aren't onto something with the same row sharing power output as it does make sense.

 

[BallistixOnZ490]

I think OP mentioned it was an RM1000x though so not really old nor unfit for high power GPUs.

Agree that no one should take any risks with PSUs, and that includes only using high quality units and never daisy chaining cables. I tried googling around to find info on rails but there really isn't any info out there, I'm hoping you aren't onto something with the same row sharing power output as it does make sense.

It was just a maybe and nothing more, its far more likely that just the one PSU was faulty.

But - all these PSUs are rated however highly they are based on pre RTX 3080 hardware, remember that there aren't many, if any PSUs that have been rated since and specifically for running these new GPUs.

And even if they do review current PSUs and they work fine initially, who is to say if these GPUs aren't putting too much strain on them? Suddenly that 10 year warranty on my EVGA T2 is more like only good for 3 years once I connect it to a 3080 because it was never tested to be viable for running such a GPU for 10 years.

 

[EsaT]

It was just a maybe and nothing more, its far more likely that just the one PSU was faulty.

But - all these PSUs are rated however highly they are based on pre RTX 3080 hardware, remember that there aren't many, if any PSUs that have been rated since and specifically for running these new GPUs.

You're pretty much out like snowman when it comes to electric engineering and just random shotgunning to every direction hoping something hits, aren't you?

PSU doesn't give out more current/power than what connected load draws and can't be the cause of fire that way.

Again excessive output voltage would require very unlikely double failure of both voltage control breaking down and over voltage protection not working.
And enough overvoltage to start fire from too high forced current would start frying things everywhere and not just in single spot.
In fact it would likely kill VRM MOSFETs (propably literally blow them up) followed soon by capacitors.


Neither has ampere of current changed from what it was in the past and it doesn't depend on what kind load configuration is drawing it.
RM1000x was definitely build to power dual graphics card configs drawing more than single 3080.

Also everything in it is likely working perfectly.
That size PSU simply isn't capable to noticing average short any more than towing truck would notice locked up breaks in car it's pulling.
12V line over current limit is likely around 90-100A to tolerate transients.
And exceeding such current demands whole current loop resistance to be almost only 0,1 ohms.
That doesn't leave really any room for resistance of wires and connectors to not limit current below triggering point.

And for perspective even half of that current would create enough heating power in shorted spot to start fire.
In focused spot without good paths for heat to disperse just sub 20A would easily create high enough temperature in seconds for something to start burning.


In short everything in this case is consistent with either slot failing mechanically and shorting, or VRM of graphics card having some failure and drawing way too high current from slot's contacts.
(these Nvidias have very high slot current draw already normally)

 

[BallistixOnZ490]

You're pretty much out like snowman when it comes to electric engineering and just random shotgunning to every direction hoping something hits, aren't you?

PSU doesn't give out more current/power than what connected load draws and can't be the cause of fire that way.

Again excessive output voltage would require very unlikely double failure of both voltage control breaking down and over voltage protection not working.
And enough overvoltage to start fire from too high forced current would start frying things everywhere and not just in single spot.
In fact it would likely kill VRM MOSFETs (propably literally blow them up) followed soon by capacitors.


Neither has ampere of current changed from what it was in the past and it doesn't depend on what kind load configuration is drawing it.
RM1000x was definitely build to power dual graphics card configs drawing more than single 3080.

Also everything in it is likely working perfectly.
That size PSU simply isn't capable to noticing average short any more than towing truck would notice locked up breaks in car it's pulling.
12V line over current limit is likely around 90-100A to tolerate transients.
And exceeding such current demands whole current loop resistance to be almost only 0,1 ohms.
That doesn't leave really any room for resistance of wires and connectors to not limit current below triggering point.

And for perspective even half of that current would create enough heating power in shorted spot to start fire.
In focused spot without good paths for heat to disperse just sub 20A would easily create high enough temperature in seconds for something to start burning.


In short everything in this case is consistent with either slot failing mechanically and shorting, or VRM of graphics card having some failure and drawing way too high current from slot's contacts.
(these Nvidias have very high slot current draw already normally)

Then feel free to explain at such if you know so much, theres no need to be so rude over it.

Funny how mad anyone here gets over things they simply disagree with.

Even the best PSUs can still blow, or even if if was the motherboard slot that too.

Minimising this kind of risk requires extremely careful selection of computer parts to be paired with these new GPUs, here is was a decent RM1000 PSU sure, and what about all the people that are still advising and going ahead and pairing a 3080 with any old 750w PSU so long as it has two PCI-E cables?

I would rather that no one elses PCs blow up while using a 3080, and if I manage to get one mine as well. Minimizing any risk for this is all I am trying to discuss.

 

[Admetos]

Something related to this thread that might help everyone.

All my computers and HiFi gear, have an RCD between the wall socket and the power strip.

I do this for 3 main reasons

1) My power strips are tacima surge / noise filters, it's very unlikely but still possible they could catch file if they fail.
2) I sometimes have equipment running at night when I'm sleeping.
3) If there is a short on any equipment, including the PC, the wall RCD should react faster then the RCD on house electrics.

4) As an extra bonus, if there is an external surge in the electric, such as a thunder storm effecting the mains, the wall RCD will trip breaking the circuit. I've seen this actually happen during a thunderstorm, how much in reality this would protect my electrics I can't say, but it did trip and break the circuit.

These are the things I'm referring to. Admittingly it sounds a bit OTT to have these all around the house, never the less it's what I do!

https://www.toolstation.com/rcd-ada...XcJ-BAJLUb0HSfOKq-RoCwyIQAvD_BwE&gclsrc=aw.ds

So there is some food for thought, if you have computers left unattended, consider one of these RCD trips between your wall socket and power strip!