News Technician Repairs Hundreds of RTX 4090s With Melted Power Connectors Every Month

[TJ Hooker]

When Nvidia came out and said it was user error that caused the melting they basically gave AIBs a blanket to deny fixing melted cards under warranty. Physical damage is never covered under warranty unless it's proven to be a manufacturing defect. Also using third-party or modded cables voids the warranty.

Exactly, which means they're most likely blocking people from using their warranty like they should and not losing money out of a scenario where people would get their GPUs replaced under warranty if they hadn't said that. Sneaky bastards.

Do you have any evidence of OEMs denying warranty on melted connectors, where 3rd party adapters/cables were not used?

And even with 3rd party adapters, I don't know if any OEMs have come out and said they won't honor warranty if they're used. All I've seen was just one guy on reddit who said his claim was denied because of that.

 

[TJ Hooker]

There were problems that didn't happen with older connectors. Then they also said you can't bend the cabling too much near the connector, to the point that, if I recall correctly, it was mentioned that you might want to check if your case has enough extra side clearance to accommodate this connector's needs.

Same thing applies to all PC power connectors. The 6/8 pin connectors also specify a minimum length of cable near the connector that shouldn't be subjected to sideways force/bending.

 

[King_V]

Same thing applies to all PC power connectors. The 6/8 pin connectors also specify a minimum length of cable near the connector that shouldn't be subjected to sideways force/bending

Not to this degree, though. The new connector disqualifies a lot of cases that wouldn't be disqualified with the existing 6/8-pin connectors

 

[TJ Hooker]

Not to this degree, though. The new connector disqualifies a lot of cases that wouldn't be disqualified with the existing 6/8-pin connectors

The length for 8 pin connectors is 19.1mm, the length for 12vhpwr is 30mm. So sure, there's a 1cm window where an 8 pin would work but a 12vhpwr wouldn't.

 

[JeffreyP55]

NorthridgeFix repairman says that he has to repair hundreds of GeForce RTX 4090 graphics cards every month and says the 12VHPWR connector failures are not a user error.

Technician Repairs Hundreds of RTX 4090s With Melted Power Connectors Every Month : Read more

The techie loves to inhale ozone early in the morning. Don't buy that garbage. On the plus side, it keeps PC repair dudes in business.

 

[JeffreyP55]

NorthridgeFix repairman says that he has to repair hundreds of GeForce RTX 4090 graphics cards every month and says the 12VHPWR connector failures are not a user error.

Technician Repairs Hundreds of RTX 4090s With Melted Power Connectors Every Month : Read more

What gauge wires are used and max "I" (current) of the GPU? Looks simply like wrong gauge wire. Obviouly generate tons of heat, wasted energy. Too cheap to do the right thing?

Common US Wire Gauges – AWG Gauges Vs Current Ratings - Resources For Electrical & Electronic Engineers

The AWG – American Wire Gauge – is used as a standard method denoting wire diameter, measuring the diameter of the conductor (the bare wire) with the insulation removed. AWG is sometime…

electricalengineering123.com

 

[ilukey77]

ive lost count of the amount of times ive pulled my 7900xtx out of my system or rebuilt my system in a new case or to just change something ..

in fact i just pulled the gpu out few days ago to swap my cpu cooler from the NZXT x63 280mm back to the Noctua NH-U12a air cooler and in a week or so im going to be swapping that to a themalright phantom sprit 120 evo to see what its like ..

may even change from the jonsbo d31 to the cooler master qube 500 all black with some custom paint work

Point is the 12vhpwr plug is messy at best the old pcie plugs are far safer and user friendly ..

how many times on average do you put your 4090 ( or any gpu for that matter) in your case and never take it out again pc builders are always playing around with stuff upgrading etc etc ..

User error maybe but the true cause is Nvidia went with a flawed power delivery system then blames it on user error because you should only build your system once and never upgrade or never DIY pc build !!

I bought my 7900xtx red devil over a 4090 for 2 reasons ive never had annoying issues with AMD gpu's ( few little issues here and there that were fixed over time but nothing for me to throw it out the window ) and cost the 4090 is and was just to crazy priced in Australia ..

The melting plugs scare me for the fact i dont want to stuff around with downtime and sending it back repairs etc etc ( not to mention the potential fire hazzard !! ) and im always pulling my gpu to upgrade , tweek , change parts in my pc ..

I rather not have the worry in my mind by using the 12vhpwr cable !!

 

[King_V]

The length for 8 pin connectors is 19.1mm, the length for 12vhpwr is 30mm. So sure, there's a 1cm window where an 8 pin would work but a 12vhpwr wouldn't.

And that was enough of a change to disqualify many otherwise usable cases.

Aside from that, it was one of multiple issues, necessitating a need for the design change.

This is not user error no matter how much Nvidia would prefer to say otherwise.

 

[truerock]

The article doesn't provide information in regard to what is causing the problem.

 

[pf100]

It's quite impressive at the number of nvidia fanboys here that refuse to believe that their God Jensen could possibly be at any fault whatsoever and it simply must be user error!

 

[schwaggins]

well ill do a small rant about him

use an powered solder sucker, a manual one will kick on the pcb when it does a kick back, for sure he take the big approach to heat and desolder the plug at once, the ground plane is massive on many cards

you have to go fast

but theses connectors melting is a bad design at 1st, power pushed to the limit on thesess pins i would say

yeah tons of discutions already, but i stand on a bad design at start

yeah theorithical power limit to the pins blah blah you never push 80-100% current in pins the margins are to narrow

Powered solder suckers are entirely unsuitable for this job what he should be using is one with a larger chamber & silicone tube on the tip, that thing in the video is for kids you can see him struggling with it

 

[jeffy9987]

are there any reports whatsoever of 4080 s buring?

 

[TJ Hooker]

And that was enough of a change to disqualify many otherwise usable cases.

Aside from that, it was one of multiple issues, necessitating a need for the design change.

This is not user error no matter how much Nvidia would prefer to say otherwise.

But the design change only affects improperly seated connectors, which is a result of user error...

Edit: I'm not saying there is 100% no design issue. But the 'they made a design change, which proves it wasn't a matter of user error' argument doesn't have me convinced.

 

[Greg7579]

Could someone please tell me if I buy a new ASUS 4090 Strix (waiting list), will it 100% for sure have the new 2x6 connector? I heard everyone was changing over on every new 4090 (especially ASUS). But is it so?

 

[willgart]

I’d like to have heard more of his reasoning for why its a design flaw and not user error.

I’m not sure why it coudln’t be both. A connector that makes it exceedingly difficult to actually seat properly is a design flaw - but one that can be mitigated by extra caution by the user.

If he’s only getting burnt cards and not dismantling the builds, there isn’t really any way he can know for certain these connectors were “properly seated” when the damages occurred except to take users’ word for it.

check the video rambling about the 12VHPWR failures
from actually hardcore overclocking.
you'll understand why it's a design issue.

 

[willgart]

the real problem is a GPU that uses too much power

not this one.
the 4090 is consuming less power than expected, and this new connector not required at all
a standard power system would be enough.

 

[YSCCC]

But the design change only affects improperly seated connectors, which is a result of user error...

Edit: I'm not saying there is 100% no design issue. But the 'they made a design change, which proves it wasn't a matter of user error' argument doesn't have me convinced.

It is a design issue as it’s too easy to unseat, if that’s only user error your cars will have no airbags, no ABS, no power steering and no reverse cam or radar… power sockets won’t have earthing pins… a bad design isn’t one that will go in smokes and flames all by itself, that is a stupid design that would never see the light of the day. But a bad design is one that is prone to be mishandled and causing catastrophic failure. Say a IT server allowing the receptionist to one click delete all data without backup, the event is Mis-clicking by the receptionist, but the root is flawed server authentication setting design by the IT department, if anyone have to be fired, it’s the IT guy not the receptionist.

 

[LabRat 891]

Every single person with an iota of "hands-on experience" dealing with low-voltage, high-current, *knew* this was a terrible design.

Even 'by-spec' there's virtually no safety margin:
Previously (w/ 6/6+2/8-pin) 50+% over specification wouldn't (typically) cause a problem.

IMO, There were at least 3 entirely different ways to move beyond 150W-rated 8-pins.
Making a connection standard with more smaller pins, seemed like a bad idea, from the start.
-based on how every other industry (except EVs) designs connectors for such applications.

 

[schwaggins]

After researching it's quite clear this is not just user error the base specification and CEM 5.0 is inconsistent.
Poor spring contacts(3 dimple astron), no strain relief to prevent uneven contact resistance from lateral load, unspecified connector retention force, no last pin engagement for sense/sideband, even NVIDIA provided faulty connectors.

the fact that it burns if it's bumped or slightly moves is unsurprising it's designed this
way

 

[SyCoREAPER]

Im not saying he's lying but i don't trust him. He's cocky, he's arrogant and he exaggerates everything to dramatic effect in his videos.

Not a fan.

Where are the reddit posts for example? News articles? There was one recently about that guy that supposedly owned it for a year and it failed. His story wasn't adding up.

This whole single source non-vetted news posting here needs to stop. It's starting to feel like a tabloid site.

 

[King_V]

But the design change only affects improperly seated connectors, which is a result of user error...

Edit: I'm not saying there is 100% no design issue. But the 'they made a design change, which proves it wasn't a matter of user error' argument doesn't have me convinced.

That's not the argument. The initial design of the new connector caused notably higher rates of failure than the old 6/8 pin. Further, a few relatively simple changes to the new design resolved some very obvious shortcomings in that initial design. The changes are not limited to helping just with improperly seated connectors. That said, your assertion that improperly seated connectors are automatically user error is false. The initial design is bad enough that improperly seating can more easily happen.

 

[spongiemaster]

I wouldn't pay this dude too much mind. He's just out to drive clicks to his site and promote himself. He's been posting 4090 videos for a while and been called out by Gamers Nexus for misleading data in his videos. Most of his videos just look like informercials for himself. Everyone else is wrong, everyone else does terrible work. Only I know the truth and how to fix things. One of them he actually holds up a 4080 PCB and calls it a 4090. He wants to call out everyone else but can't identify a 4080 vs a 4090 PCB that he's holding?

 

[lynn_levi]

theoretically speaking, the 16 pin connector is basically 2 smaller 6 pin connectors merged into one plus 4 extra sense pins; obviously the problem still lies on the connector, it is small for how much power it is meant to carry over, ~600W;

think of it like blowing a water current at 15000psi through a tube joint connector which has just about, 4cm of diameter (i aint a good physician but this is example), the joint connector will be BLOWN FROM INSIDE as the smaller water density being limited by the connector also massively increases the internal pressure in exchange;

same will happen to the 16 pin connector, ~50A of DC current being blown through such a connector plus smaller metal contacts (lower resistance) will burn the connector over time. this is not even a design flaw, this is a bad decision in the first place for a new connector meant for heavy power delivery

the particular user error would be not inserting the plug fully in as the connector will constantly arc (try shaving a live 220VAC cable with a knife) and as well burn the connector from inside; this can be totally avoided but such a small connector like that for high power delivery is a fire hazard by itself, because no electrician in their right mind would use a 14AWG cable for 5+kW appliances. should have sticked with the older but bigger 6+2 pin connectors. two 6+2 pin connectors will do the same job just fine. 900W? one more connector. sticking with this 16 pin connector while being afraid of fire? add one more 16 pin connector. come on AIBs

 

[SeaTech]

Let me give my two cents as someone who has recently bought a new 4090 that has the old connector (non-founders). To be clear the original design and execution was flawed, period. Upon installation I noticed quite clearly that it will feel like it is in all the way with relatively little effort- but the connector is NOT actually all the way in! Jay z two cents and gamers nexus did some good videos showing the issue...but a user probably won't appreciate just how much force is need to get the connector in ALL the way. It took an extraordinary amount of effort just to get that extra 10-20% in. So I can see very easily how someone not looking closely enough at the connector head or in poor lighting puts it in 80% of the way and maybe bends the connector cable trying to shut the case door (some 4090s are simply too wide for a regular mid tower case) could potentially run into trouble later with that connector backing out/slipping out further...and then in a full load scenario you end up with real problems. You have to ask: is it really the user's fault if the connector feels like it is in when it is NOT. Or that it requires an unusual amount of force to get 100% of the way in? Or that the 10-20% that sticks out is not immediately obvious to the user? They could have at least created a big neon or red mark on both sides of the connection indicating correct placement.

Nvidia needs to help get the connector fix out to market for ALL of the AIB non-founders cards or this problem will continue to happen. "User error" is not an acceptable excuse for improper design execution.

 

[lynn_levi]

What gauge wires are used and max "I" (current) of the GPU? Looks simply like wrong gauge wire. Obviouly generate tons of heat, wasted energy. Too cheap to do the right thing?

Common US Wire Gauges – AWG Gauges Vs Current Ratings - Resources For Electrical & Electronic Engineers

The AWG – American Wire Gauge – is used as a standard method denoting wire diameter, measuring the diameter of the conductor (the bare wire) with the insulation removed. AWG is sometime…

electricalengineering123.com

i believe wire gauge is barely a contributing problem. and i believe 16 pin connectors use 16AWG cables