News Nvidia power meltdowns hit 8-pin adaptors — old power supplies are the common factor

Toggle sidebar Toggle sidebar
Going forward for RTX 60x0, this is the proper power connector for an Nvidia GPU since they are so power hungry.

The old tried and true. This one aint gonna melt. ATX24.

MICRO-ATX-3__94745.1559919434.jpg
 
  • Like
Reactions: adbatista, TCA_ChinChin, NinoPino and 4 others
It may be 15 years old but at north of 800w and factoring in 2% loss per year, it's still capable of over 600w, which is twice the power draw of the 5080 (as measured by TH in the 5080FE review) and sufficient for the system. It also should not have melted the connectors as the PCIe 8 pin is capable of well over 150w of power before failing (over 200w I believe).

Even though this person demonstrated...questionable judgment, to be kind, in using a 15 year old PSU in a gaming machine, it still shouldn't have happened unless there was physical damage to the connectors, or to the adapter cable.
 
  • Like
Reactions: TCA_ChinChin, SydB, NinoPino and 2 others
edzieba said:
Alvar Miles Udell said:
It may be 15 years old but at north of 800w and factoring in 2% loss per year, it's still capable of over 600w, which is twice the power draw of the 5080 (as measured by TH in the 5080FE review) and sufficient for the system. It also should not have melted the connectors as the PCIe 8 pin is capable of well over 150w of power before failing (over 200w I believe).

Even though this person demonstrated...questionable judgment, to be kind, in using a 15 year old PSU in a gaming machine, it still shouldn't have happened unless there was physical damage to the connectors, or to the adapter cable.
Click to expand...
Click to expand...
It happens because 4080, 4090, 5080, and 5090 can't load balance ANY of the positive 12V pins; the connector type, then, matters less as whether it's 12VHPWR, 12V-2X6, or multiple molex 8-pin connectors, too much current on a single pin / wire is going to allow this sort of thing to happen.

As others have said, having fewer larger-gauge wires would provide immediate relief to this problem, even if it doesn't completely eliminate it as again, it goes back to the power design flaw of these GPUs that nVidia continues to fail to admit.

Also, I'm in the same pool of thinking that old but high-end, high-power PSUs have never been a true problem in the past and shouldn't be now, but green changed that. This article mentions newer PSUs (ATX 3.0 and especially 3.1) having better safety mechanisms, but they also don't load balance +12VDC on those pins, so it's kind of a moot point IMO if they aren't stopping this fiasco from happening. If anything, this PSU demonstrates its quality and indeed, power output loss over time is really the biggest issue of any, and that still isn't a big one given the hardware that's in this PC.
 
  • Like
Reactions: adbatista, dalek1234, bigdragon and 2 others
There have been comments regarding Nvidia graphics cards "load balancing" across its positive power pins/wires. That would be a bad idea.

It would be great if Nvidia graphics cards monitored the volts and amps in the individual pins and wires and report that info back to the operating system. If the individual pins/wires have something different on them than 12 volts and/or something more than 8.4 amps (plus/minus a few watts per wire/pin) the operating system should shut down - it should not "load balance out-of-spec pins/wires.

OK... so I'll be more specific about amps. Yes, 8.4 amps would be the expected amount if the draw is 600 watts. Obviously, less if the graphics card wants less electricity. The issue is, if there is an unbalanced amount of amps across the positive pins/wires then, ideally, that would be reported to the operating system. And, also, each plug on a PC PSU is rated for a maximum safe amount of amps on each individual wire. If that maximum is exceeded on any wire/pin, it ideally would be reported to the operating system.
 
Last edited:
  • Like
Reactions: dalek1234, NinoPino and artk2219
dwd999 said:
It just amazes me that users will spend a huge amount on a gpu and/or cpu and then cheap out on a power supply or some other component like the motherboard, cooling fans, etc. that's just as important as the gpu and/or cpu.
Click to expand...
That PSU was not "cheap", and a new PSU of same category is very expensive.
From a economic and ecological pov he did the right choice.
 
  • Like
Reactions: bigdragon
NinoPino said:
That PSU was not "cheap", and a new PSU of same category is very expensive.
From a economic and ecological pov he did the right choice.
Click to expand...
Based on it being 15 years old, I wouldn't be surprised if he got it from a friend for free when they got a new one. That would qualify as being cheap. If you're spending over $1k on a GPU, using a 15 year old PSU to power it is not the right choice, it's dumb.

That card comes with a THREE 8-pin connector splitter. He claims he was using all three, but the third one didn't melt which allowed him to unplug it for the picture.
 
  • Like
Reactions: dwd999 and bolweval
It's still a defective connector. I'm sure Nvidia is following PCIe specs, but whoever at PCIe responsible for designing this connector did not follow the best practices for designing a connector used for high current applications like this.

A connector like this wouldn't be allowed to be used OUTSIDE the computer for anywhere near that much power. You'd have electrica outlets melting and electrical fires all the time if they were designed so they run 100C+ in perfect conditions and melt if the plug is just the tiniest bit loose. But they have margins in wire size, margins in the amount of metal to metal contact, design to ensure the outlet isn't 'lose'', etc. So when you DO have an outlet melt it's because you're running at power limit (like a space heater) AND the plug is loose AND you have worn out outlets (outlet is 'loose', possibly corrosion and wear reducing metal to metal contact.)
 
spongiemaster said:
Based on it being 15 years old, I wouldn't be surprised if he got it from a friend for free when they got a new one. That would qualify as being cheap. If you're spending over $1k on a GPU, using a 15 year old PSU to power it is not the right choice, it's dumb.
Click to expand...
So, using a perfectly working hardware, is dumb only because it is old ?
Really ?
For me, transforming that hardware in e-waste, is dumber.
 
  • Like
Reactions: tumbbi
bolweval said:
On a 15 year old power supply, i would imagine that 8 pin connector had probably been plugged in and out 100's of times, it's load capacity would have been reduced I'm sure..
Click to expand...
Why you assume that the cables are unplugged and plugged once a month or more? I have 11 years old PSU and the cables are max ten times unplugged and plugged. This PSU has seen three different GPUs, so I don't update GPU often but still, I cannot see why anyone would get that high count.
 
  • Like
Reactions: obvious oscar and SydB
It's puzzling to me why they don't just go to specialty gpu supplies with 24 or 48 volt outputs to drop the current. Solves the issues and also allows the marketing of another product. They're at the limit for this voltage.
 
Usually when i do a major upgrade--i make sure to upgrade the Power supply to support the newer Components or if i take a pre configured build say from my local shop or a business. I upgrade the Power supply as soon as i can afford to.

I typically don't trust Power supply included for too long

Bought an Ibuypower Ready Built in July 2024, so far i stayed with the Power supply it came with, but i do plan on upgrading it soon as i can for peace of mind, better quality, and make sure the components last as long as possible

Right now it has a 750 Watt ATX Gold Ceritified HighPower Power supply (exact model i'll have to check on when do the next dust clean out)

((Not Too capable of building myself, but i've done Power supply upgrade on my old Asus PreBuilt, and my old Intel 10700 local computer store build) so that part im comfortable with))

Previous system i chose an EVGA G3 650 Watt back in 2020

Might go Corasir when i choose next one or
Seasonic maybe, or maybe another EVGA

***Sometimes i feel ashamed that i don't got the Real-Life Peace and Quiet to Build my own Fully, so typically i don't even mention its an Ibuypower Ready Build***
 
Ironically, to get automatic foolproof failure-proof load balancing to prevent connector overheat, all you need to do is... use thinner cables......
 
DS426 said:
It happens because 4080, 4090, 5080, and 5090 can't load balance ANY of the positive 12V pins; the connector type, then, matters less as whether it's 12VHPWR, 12V-2X6, or multiple molex 8-pin connectors, too much current on a single pin / wire is going to allow this sort of thing to happen.
Click to expand...
^^^ This...
...which makes this whole article a nothing-burger. 4000/6000 is still to blame
 
  • Like
Reactions: daeros
tumbbi said:
Why you assume that the cables are unplugged and plugged once a month or more? I have 11 years old PSU and the cables are max ten times unplugged and plugged. This PSU has seen three different GPUs, so I don't update GPU often but still, I cannot see why anyone would get that high count.
Click to expand...
I assume they have been, you assume they haven't, both ASSumptions. ; )
 
  • Like
Reactions: obvious oscar
bolweval said:
On a 15 year old power supply, i would imagine that 8 pin connector had probably been plugged in and out 100's of times, it's load capacity would have been reduced I'm sure..
Click to expand...
Plugged in and out hundreds of time!

What the heck dude 🤣 I only install a new graphics card every 3 years roughly.. you doing it weekly ?
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom