First report of an Nvidia RTX 5080 power connector melting emerges

[midian182]

What just happened? With several reports of RTX 5090 power connectors melting, it's starting to look as if the dreaded issue that was found by many RTX 4090 owners has made it to the Blackwell generation. In what could be even worse news for Nvidia, the first incident involving an RTX 5080 has surfaced.

Redditor Ambitious_Ladder1320 posted images of a 16-pin power connector that had melted on the PSU side. The power supply unit in question was an Asus ROG Loki, which was plugged into an RTX 5080.

According to the post, the user saw a red light flashing on the RTX 5080, indicating that a pin wasn't seated properly, despite both CPU and GPU cables being locked into place.

Ambitious_Ladder1320 then powered down the PC and reconnected the cable. Upon restarting, the warning disappeared, but the resolution and the refresh rate had dropped and the GPU reported being switched to PCIe 3.0 mode.

After shutting down for a second time and reinserting all the cables, it was discovered that the connector on the PSU end of the 16-pin 12VHPWR cable had melted. The GPU end and the GPU itself appear unaffected.

Does Rog Lokis molted rtx 5000 gpu 12vhpwr cable
byu/Ambitious_Ladder1320 inASUS

The user said he had seen three similar reports from ROG Loki PSU users on YouTube, though he never specified if they were also using RTX 5080 or RTX 5090 cards.

It appears that we'll soon know more details about what happened as Steve Burke from Gamers Nexus asked if the YouTube channel could buy the GPU, cable, and PSU for testing and RMA support evaluation.

There were reports earlier this week of an RTX 5090 FE cable's connector melting on both ends and damaging the card and PSU. It was a third-party cable from Moddiy, though the owner said he'd used it for years with his RTX 4090 and experienced no problems.

Der8auer replicated the user's setup using his own components – though he used a Corsair 12VHPWR cable – and ran the Furmark test. The cable's connectors reached 150°C on the PSU side and close to 90°C on the GPU side. The problem was an uneven distribution of power: two wires designed to carry 5 to 6 amps of current were carrying more than 20 amps each, while some cables carried as little as 2 amps.

Permalink to story:

First report of an Nvidia RTX 5080 power connector melting emerges

 

[yRaz]

This is what happens when the bean counters in accounting tell the engineers what to do. The profit margins on these cards is already absurd, what do they have to gain by saving a few pennies designing a circuit that uses few parts? They aren't even selling enough cards for the savings to add up to cover the salary of one of the bean counters.

They shouldn't be trying to save money on something that is being priced and sold as a luxury item. I don't have a problem with $1000-2000 graphics cards, just make it a good product that provides $1000-$2000 worth of value. Part of a luxury experience should be peace of mind that you made a good purchase.

 

[wiyosaya]

Surprise, surprise. /s

 

[redgarl]

So, it was all related to the amount of power going through the cable in the end.

No "USER ERROR", just design of the PCI-SIG standard.

 

[GGG777]

This is what happens when the bean counters in accounting tell the engineers what to do. The profit margins on these cards is already absurd, what do they have to gain by saving a few pennies designing a circuit that uses few parts? They aren't even selling enough cards for the savings to add up to cover the salary of one of the bean counters.

They shouldn't be trying to save money on something that is being priced and sold as a luxury item. I don't have a problem with $1000-2000 graphics cards, just make it a good product that provides $1000-$2000 worth of value. Part of a luxury experience should be peace of mind that you made a good purchase.

"This is what happens when the bean counters in accounting tell the engineers what to do. The profit margins on these cards is already absurd, what do they have to gain by saving a few pennies designing a circuit that uses few parts?"

Really???
Look that 20A and not instant melting the cable so this is quality part with an unintended use.
The FurMark NOT a intended use.

With under playing game Star wars outlaw working fine.

 

[Jonny5]

This happened to another chap without running furmark tho how do you explain that one

 

[kiwigraeme]

Good thing they sold so few. especially in EU, Australia and NZ

they would be hammered in consumer protection

user error has to be quite egregious and average user is known to do silly things

eg remove coins from pocket before washing clothes. Sorry washing machines should survive such a common mishap, if you have a n outer plastic shell and coin gets in there due to bad tolerances and puts a hole in it , manufacturer pays

These cards are knowingly going to be powered by cables coming from established PSU manufacturers.
Given it's always considered best practice to us PSU cables to stop any mix up in wiring etc

Nvidia cheaping out and not wanting to freak people out about how much power really needed.

Really these need PSU extra gpu/Aux cables to attach to end of GPU as well a- all same 12v as existing I think, Nvidia is cheapening out as needs more hardware/software controllers

 

[yRaz]

Good thing they sold so few. especially in EU, Australia and NZ

they would be hammered in consumer protection

user error has to be quite egregious and average user is known to do silly things

eg remove coins from pocket before washing clothes. Sorry washing machines should survive such a common mishap, if you have a n outer plastic shell and coin gets in there due to bad tolerances and puts a hole in it , manufacturer pays

These cards are knowingly going to be powered by cables coming from established PSU manufacturers.
Given it's always considered best practice to us PSU cables to stop any mix up in wiring etc

Nvidia cheaping out and not wanting to freak people out about how much power really needed.

Really these need PSU extra gpu/Aux cables to attach to end of GPU as well a- all same 12v as existing I think, Nvidia is cheapening out as needs more hardware/software controllers

well it looks like the cause of the problem is that all of the power is connected to the same rail without any balancing circuits so if there are 3 or 4 bad connections, all the power goes over the remaining cables. The card has no way to not ask for more power and the power supply doesn't realize that all 600watts is going through 1 or 2 cables. And, for whatever reason, it seems to be the same pin melting in all the connectors. For whatever reason, the same wires stop transmitting power and all going through that one pin which, to me atleast, throws a redflag for a design error in implementation on the cards side. If this is happening to people who were running the same PSU with the same power connector on a 40 series card and a 50 series card shows near identical failures with different PSUs, it's kinda hard to blame the PSU at that point. Whatever "fix" they implemented has done nothing or worse, regressed. The 5090 that made the news recently, the guy said he was running the same PSU and cable with a 4090 for nearly 2 years with no issues. He plugs a 5090 in and a week later it gets fried.

 

[user556]

well it looks like the cause of the problem is that all of the power is connected to the same rail without any balancing circuits ...

That's a red-herring. If the plug/cable can't keep the resistance even then it's a lost cause.

The fault is entirely with the plug on the cable. The flawed 12VHPWR cables should all be recalled and replaced with 12V-2x6 cables free of charge.

 

[user556]

How the plugs might have been made so badly in the first place, I have no idea. Was it one model of plug? Was it a whole range from one manufacturer? Or was it a total crap shoot for the cable makers? It's maybe a case of the cable makers hadn't needed to test the plugs because the plug manufacturers' specs had all been reliable up till now. Doesn't matter a lot to the end users. There is a mess and it needs fixed.

Part of the fix is to clearly delineate and state existence of correctly functioning 12V-2x6 cables. Providing confidence that, moving forward, the end user is purchasing reliable cables again.

 

[yRaz]

That's a red-herring. If the plug/cable can't keep the resistance even then it's a lost cause.

The fault is entirely with the plug on the cable. The flawed 12VHPWR cables should all be recalled and replaced with 12V-2x6 cables free of charge.

I don't think you understand, the 20 and 30 series had 3 balancing circuits going into the card each connected to 2, 12V+ wires. If there was a problem on any of those rails, the card would just not turn on. The 4090 had 2 balacning circuits each connected to 3 wires. The 5090 and 5080 have a 6X1 configuration. On the 20 and 30 series, wires couldn't carry more than 200 watts before the card just wouldn't shut off. On the 4090, upto 300 watts could go through 1 wire if the other two failed on the same circuit. On the 5090, you can have up to 5 of the 6 power connectors fail and it will try to send 600W over 1 wire and the card will still turn on.

 

[user556]

I don't think you understand, the 20 and 30 series had 3 balancing circuits going into the card each connected to 2, 12V+ wires. If there was a problem on any of those rails, the card would just not turn on.

Still a red-herring. Less costly maybe, but not working is still no good. The cable is still what has to be fixed.

 

[yRaz]

Still a red-herring. Less costly maybe, but not working is still no good. The cable is still what has to be fixed.

That's not a red herring, that's litteraly why they are melting. The Connector is a **** connector, you'll never get any argument over that from me. The fact is that the wiring on the board makes the cards more likely to send more power over a single wire instead of spread out across the entire connector. Nvidia didn't need to make a **** connector and take away MORE safety features.

 

[user556]

The card can't operate that way. The only option is for it to shut down due to imbalance. That's good protection when there is a problem but it doesn't solve the problem. Hence it's a red-herring.

And Nvidia may have had input in choosing which connector type to go with but they certainly didn't design nor make them. You need to look at outfits like Molex for that.

 

[yannus]

Absurdly priced, absurdly designed item for absurd people.

 

[GGG777]

This happened to another chap without running furmark tho how do you explain that one

Obviously, who will take responsibility?
Interestingly, there are no issues during normal testing, be it gaming or benchmarking. Only FurMark.

 

[yRaz]

The card can't operate that way. The only option is for it to shut down due to imbalance. That's good protection when there is a problem but it doesn't solve the problem. Hence it's a red-herring.

And Nvidia may have had input in choosing which connector type to go with but they certainly didn't design nor make them. You need to look at outfits like Molex for that.

There is a reason breaker boxes exist. It's so that if there is an issue with the wiring, there is a safety in place to prevent further damage, like a house fire. Now, we even put GFCI outlets on they other side of the wiring. Now, the wiring in the house is spec'd to avoid issues, but they but breaker and GFCI outlets inplace to STOP further damage before it had a chance to occur. What nVidia did was the equivalent of removing a breaker and GFCI outlet. It'd be one thing if they NEVER did that, but they spec's previous generations of cards with those safety measures in place. In the 40 series, they removed one of those safety measures, going from 3 balancing circuits to 2. This caused melting problems. Knowing full well that this was a problem, the reduced the number of circuts from 2 to 1 and increased the power going over the cable.

They used a bad cable, knew it was bad after it was out in the wild and then removed EVEN MORE safety features. I'm not trying to downplay the issues the cable has, it 100% is a bad connector. However, we can't give nVidia a pass for removing safety features after finding out just how bad the connector was.

We need to hold them accountable for removing safety features. They should have used more safety features, not less. We are talking about fractions of a percent of the over all cost of the card. For the money they're saving, the savings wouldn't even add up to the cost of a single 5090 relative to the amount of units they've sold

 

[user556]

Maybe that can be classed as safety. I don't think anyone has complained of burning down their house just yet.

It's still not fixing the real fault though. A circuit breaker protects the building, not the machine. The machine is still out of action because there is a fault with it. In our case that fault is the flawed cable plug.

PS: I'm not against having the protection. It's a great feature. But the plug still is the actual problem that needs fixed. And by all accounts the plugs on the newer 12V-2x6 (H++) cables are indeed fixed.

Now just need to dispose of all the the older 12VHPWR (H+) cables so they won't continue to cause future problems. Which should be done as a general recall programme.