Close call - reminder to buy a quality psu and check/replace if old

Associate
Joined
12 Jun 2012
Posts
86
Location
Plymouth
A moment of silence for my 7 year old XFX Pro850w

The cable leading to GPU melted itself and committed suicide (not sure why?) after a good 7 years of service. Created an awful smell... The whole house stunk of burned plastic you do not want to smell burned PSU

A reminder not to skimp on your power supply, my xfx failed and the included protection saved the rest of my rig. Also if you have an older Psu maybe it's time to check using a multimeter/replace the old girl.

photo before and after I got the pliers out, was truly plastic welded on there

 
It still shouldn't fail like that.
Tell me about it, I still can smell burnt psu in the office. It had been working with my 1080ti for some time in this config and I have made no recent changes so your guess is as good as mine.

Opted for a replacement psu with a 10 year warranty so shouldn't have to be worried about this again :)
 
That was a high quality psu and with psu's offering 7 and 10 year warranties these days it's not that old really. Just out of interest were you powering that 1080ti from the two pci-e leads that are daisy chained from a single lead at the psu? If so it could be the case that the gpu was drawing too much power down the one single cable. A 1080ti can easily pull over 300w, more when overclocked. It's always advisable to power these high end cards with two seperate pci-e leads from the psu.
 
That was a high quality psu and with psu's offering 7 and 10 year warranties these days it's not that old really. Just out of interest were you powering that 1080ti from the two pci-e leads that are daisy chained from a single lead at the psu? If so it could be the case that the gpu was drawing too much power down the one single cable. A 1080ti can easily pull over 300w, more when overclocked. It's always advisable to power these high end cards with two seperate pci-e leads from the psu.
There's decent chance of this eventually happening again, if you try to power 300W card through single cable.

https://knowledge.seasonic.com/arti...ark-for-high-power-consumption-graphics-cards

Hmm i remember thinking about this years ago... then i saw this video by jayztwocents https://youtu.be/UL7KIVI_hJg?t=340 where he compares 1 pci-e cable (+daisy) vs 2 cables.
He comments that the power limit for 1 cable should be 396w which is well over the 330w limit of my GPU so i just stuck with the 1 cable and never thought about this again.

Just to be safe I will add another pci-e connector :D
 
That was a high quality psu and with psu's offering 7 and 10 year warranties these days it's not that old really. Just out of interest were you powering that 1080ti from the two pci-e leads that are daisy chained from a single lead at the psu? If so it could be the case that the gpu was drawing too much power down the one single cable. A 1080ti can easily pull over 300w, more when overclocked. It's always advisable to power these high end cards with two seperate pci-e leads from the psu.

I don't care what internals it is based on XFX in my experience have always skimped on something somewhere and I wouldn't touch them with a bargepole personally. (See my post here for instance https://www.overclockers.co.uk/forums/posts/28349779 ). Over the years of supporting other people with IT stuff I've seen so many instances of XFX stuff and stuff going pop, etc.

Though the advice on wattage and single cable, etc. is definitely something people should take notice of.
 
Hmm i remember thinking about this years ago... then i saw this video by jayztwocents https://youtu.be/UL7KIVI_hJg?t=340 where he compares 1 pci-e cable (+daisy) vs 2 cables.
He comments that the power limit for 1 cable should be 396w
Which is wrong.
Max current for Molex Mini Fit contact is 9A.
So for three 12V contacts that would be 27A. (324W)
But that max current applies only for single contact.
When you start adding more of them into connector, that piles up heat derating max current per contact for allowed temperature rise.
Which is there for a reason.
Besides at some point weakening mechanical strength higher temperature also makes connector's plastic degrade faster than specified.
(neither would I want to push 9A through only 18 AWG wire with multiple of them bunched together)

And those are no doubt also for perfect "in spec" connect mating.
Thermal cycling usually works to loosen contact and if there's any loose initially, that loose can become problem faster.
And once contact isn't good anymore than means heating increases at same pace with resistance.
Which also worsens that thermal cycling.

And you should now know well what happens when contact heats too much...
 
Back
Top Bottom