News User Reports 12VHPWR Connector Melting From the PSU Side

[Admin]

A new user report shows that Nvidia's 12VHPWR power connectors are now melting from a connectors housed inside a Gen 3 power supply.

User Reports 12VHPWR Connector Melting From the PSU Side : Read more

 

[InvalidError]

But it's hard to believe that all of the errors were due to user error since some of these latest reports come from people who claim to be veteran system builders.

Even rocket scientists and engineers make mistakes. Just because a connector may have been plugged in correctly initially doesn't mean it didn't pull out during cable management, other not necessarily related internal work or transport/shipping.

 

[atomicWAR]

I'll bet they didn't have it fully inserted but it may be it is actually one of the few legit faulty connectors/connections. Could be the wire itself or on the PSU side if that is the case.

 

[Roland Of Gilead]

Damn. The gift that keeps on giving!!

 

[hotaru251]

I don't see this being "user error" given its entire bottom row & unlike GPU the clip on psu side is a lot more reliable.


even IF its user error..the amount of times "user error" happens should be a red flag to re-design the connector.

 

[InvalidError]

even IF its user error..the amount of times "user error" happens should be a red flag to re-design the connector.

How many out of how many?

Nothing will get done about it if the field failure rate is something like 0.1% - that is considered very low by consumer electronics standards. If you are hoping for a safety recall, 94V-0 plastics cannot sustain a flame, so you aren't going to get a recall for that unless you can show standard-compliant cables being set on fire despite the standard calling for materials that should make it impossible.

 

[atomicWAR]

I don't see this being "user error" given its entire bottom row & unlike GPU the clip on psu side is a lot more reliable.


even IF its user error..the amount of times "user error" happens should be a red flag to re-design the connector.

I can still see it being user error BUT it does seem a little prone to these issues for users. In a hopefully near future revision shortening the sensing wires, making the connector tab have a louder clip and doing like MSI did and coloring the connector would be solid first steps in fixing the problem I believe.

 

[InvalidError]

I can still see it being user error BUT it does seem a little prone to these issues for users. In a hopefully near future revision shortening the sensing wires, making the connector tab have a louder clip and doing like MSI did and coloring the connector would be solid first steps in fixing the problem I believe.

The real solid first step towards fixing the issue would be to have actual remote voltage sense pins to detect total connector+cable drop so the GPU can throttle to keep total wiring losses within specs and give a warning when the connection becomes excessively bad. Ex.: if wiring losses are set to 10W max, that would allow up to 200mV of wiring droop at 50A. If the cable has 300mV of droop, then the GPU throttles down to 33A to keep total wiring losses under 10W. Even if you end up with 5W of losses concentrated on a single pin, the power/ground plane the pin connect to should be able to dissipate that without burning out the connector.

 

[rgd1101]

bring back non modular psu

 

[PlaneInTheSky]

---- -----, can we stop blaming the users

this is like Apple "you're holding it wrong" all over again

imagine regular house sockets consistently melting cables and settings people's house on fire, and the community going "user error, you didn't plug in the cable correctly", stop with the damn excuses, this is 100% the fault of the people who designed this

MOD EDIT: Profanity Removed

 

[Deleted member 2731765]

So now this becomes a BIG news topic, just because some random guy on REDDIT posted his findings (aka only 1 user), and that too this is an isolated case, since nobody else has companied about this issue on the PSU side? Hmmm, Great ! Keep them coming !

/s

 

[helper800]

As I have said again and again, unless we get hundreds of reports its not statistically relevant to manufacturers. So we can calculate a minimum amount of 4090's in the wild with steam data. In May there were 25.36 million steam users. May's steam survey revealed that 0.44% of its users had 4090's. This means there are at least 111,144 4090's sold. If we then take into account 60 cases we get an occurrence rate of ~0.054% for burning connectors. Please let me know if I am wrong.

Edited for correctness, found the error I wasn't seeing.

 

[atomicWAR]

As I have said again and again, unless we get hundreds of reports its not statistically relevant to manufacturers. So we can calculate a minimum amount of 4090's in the wild with steam data. In May there were 25.36 million steam users. May's steam survey revealed that 0.44% of its users had 4090's. This means there are at least 1,111,440 4090's sold. If we then take into account 60 cases we get an occurrence rate of ~0.0054% for burning connectors. Please let me know if I am wrong.

Sounds dead on to me.

 

[chaz_music]

As power systems designer, it would cause me great concern to see the number of failures that are happening if this were my design. The typical warranty calls for most products will reflect a small number of actual field failures, as there are probably very many more and people are not reporting them. I have seen this many times over my career.

I took a look at the Molex Mini-Fit connectors that are similar to the actual one being used and found the spec for the 12 pin part. For 600W on a 12VDC system, you will need 8.33A per pin - if they all share perfectly. They won't, and a good designer plans for some redundancy to account for that. For instance, even in the NEC electrical code, they derate wires if operated in parallel. Connectors are done the same way, but with even more derating needed.

Connector vendors also play spec games and will list the maximum current per pin for a single or minimum pin housing for good marketing exposure. If you plan to use more than that 1 or 2 pins for high current, you should always derate depending upon the vendor recommendations - check the specs. This is nearly universal on power connectors and where many design engineers get into trouble when using power connectors.

In the spec chart on the Molex part, they recommend derating to 5.5A per pin with the 12 pin configuration. The derating has many factors including contact pin resistance spread (how well they will share), as well as the thermal profile of the connector housing. The pins toward the center of the housing get the most thermal stress since they get heat from both sides.

There are many design factors external to the connector that effect the reliability. One is the PCB design: did they put agressive thermal reliefs into the copper artwork? This can generate added local heating (just like using too small of a wire into a connector). Thermal reliefs are used to help with soldering and usually look like a wheel spoke pattern. Another factor is the typical humidity in the operating environment, which effects the speed of metal oxidation for the contact metals. If the connector environment also has high vibration, then the contacts can undergo fretting failure from the contacts rubbing which slowly eats the outer plating and exposes the metal underneath (usually copper). Depending upon the contact and housing designs, this might not need much vibration to dramatically accelerate this. Think about cooling fans, water pumps, and other rotating devices in a computer case.

On my past designs, I have performed thermal tests on critical power connectors like this to check the inside pins are not getting uncomfortably hot. Before the crimped pin is inserted into the housing, I would attach a thermocouple as close as I could to the contact area on the crimp pin. And choose a contact that is by the center of the connector. Bake the system until it reaches thermal steady state. The connector housing should not be anywhere close to the maximum rated temperature.

The 8.33A rating per pin needed in this application needs some inspection. And I don't believe blaming the user is a reasonable action. People are not responsible and criticized for the plug on their hairdryer for instance. It is expected to work, and safely.

 

[InvalidError]

The 8.33A rating per pin needed in this application needs some inspection. And I don't believe blaming the user is a reasonable action. People are not responsible and criticized for the plug on their hairdryer for instance. It is expected to work, and safely.

NEMA/IEC plugs are designed to get the mechanical snot beaten out of them by 500lbs gorillas for hundreds of cycles without immediately becoming a fire or shock hazard at the slightest accident and are 10X bigger than electrically necessary because of it. Internal connectors rarely have more mechanical integrity than necessary to stay in place since they are only meant to be touched by people who know what they are doing and normally tucked away behind an access panel to protect them from the environment and user(s).

 

[bolweval]

Are you telling me that we have to plug them all the way in on the power supply side too???

 

[RandomWan]

"But it's hard to believe that all of the errors were due to user error since some of these latest reports come from people who claim to be veteran system builders."

Never underestimate the potential for humans to be stupid and/or lazy.

For those who keep going on about redesigning the connector. How many total cards and power supplies have been sold? How many have experienced failure? How many of these people are truly seasoned builders?

If the failure rate is well below 1% and there's been no clear sign or proof that it's a failure of the connector, why would they redesign it? You should do some research into reliability engineering and how often you will see users using equipment incorrectly. You don't dump a bunch of money and man hours into changing something because a tiny fraction can't get it right.

The fact remains that (to my knowledge) no one has been able to replicate these failures outside of not properly securing the plug. If no one can reproduce the results outside of that failure mode, then it's not the fault of the connector design.

The 8.33A rating per pin needed in this application needs some inspection. And I don't believe blaming the user is a reasonable action. People are not responsible and criticized for the plug on their hairdryer for instance. It is expected to work, and safely.

They are if they plug it incorrectly and there are people who will do that incorrectly even if it is made relatively idiot proof with polarized plugs. They make safer equipment, nature makes bigger idiots.

 

[Findecanor]

Indirectly, this is an effect of games' hardware requirements having become ridiculously high.

Games don't need super-ultra-high resolution textures and deep pixel shaders to be playable. Graphics cards were good enough a decade ago.

I think game developers should focus on gameplay, and on optimising their games. That would make games better.

What hasn't progressed in terms of graphics quality in the last oh, fifteen years is the realism in animations and geometry. You still see the player character running over items to pick them up instead of picking them up, etc, etc.

 

[InvalidError]

Games don't need super-ultra-high resolution textures and deep pixel shaders to be playable. Graphics cards were good enough a decade ago.

High resolution textures mostly consume VRAM and VRAM bandwidth, not much GPU-power. As many have said before, higher-resolution textures are the (relatively) low-hanging fruit of increased visual quality.

For the most part though, I agree that graphics were already about as good as I can be bothered with ~10 years ago.

 

[bolweval]

---- -----, can we stop blaming the users

this is like Apple "you're holding it wrong" all over again

imagine regular house sockets consistently melting cables and settings people's house on fire, and the community going "user error, you didn't plug in the cable correctly", stop with the damn excuses, this is 100% the fault of the people who designed this

MOD EDIT: Profanity Removed

Actually, in my 40 years of being an electrician i have seen 100's of outlets burn and in a few cases burn the structure down. Some of them were electrician errors ie. forgetting to tighten a screw. Sometimes the little push connectors get overloaded an burn the wires back. Most of the time it was end users using space heaters in old worn out plugs.

It took 80 years for them to finally come up with ARC Fault protectors to detect bad connections and trip the circuit. Maybe we need ARC Fault protects for these videocard plugs?

 

[ManDaddio]

Aftet reading the comments with some being "expert input" about electricity and connectors nothing being said was helpful.
All anyone has to look at is prior connectors like was used with 3090ti.
I think what is happening with some/all of these is users not connecting the cables in all the way, overclocking and keeping silent, or bad wiring and electric systems in their houses.

 

[steve4king]

I just purchased an FSP hydro ti pro 1000w.
I haven't hooked it up yet, but looking closely at the included 12VHPWR connector, some of the leads are loose. Any pressure will push them out of the connector so that they are barely touching the device, no matter whether the connector is fully seated. I expect it would cause a fire if I used it as is. Definitely a manufacturing defect.

 

[Alvar "Miles" Udell]

There's a reason there's only three PSU manufacturers I trust: SeaSonic, Enermax, and Superflower, in that order. Not sure who makes BeQuiet since TH stopped updating the PSU manufacturer list a decade ago, but I wouldn't touch them with a parsec long pole.

 

[Deleted member 2838871]

There's a reason there's only three PSU manufacturers I trust: SeaSonic, Enermax, and Superflower, in that order. Not sure who makes BeQuiet since TH stopped updating the PSU manufacturer list a decade ago, but I wouldn't touch them with a parsec long pole.

I've never used anyone but EVGA... and have never had a PSU issue. If it ain't broke... I ain't fixin' it. The same applied to GPUs... at least until the 4000 series.

 

[Deleted member 2838871]

I don't see this being "user error" given its entire bottom row & unlike GPU the clip on psu side is a lot more reliable.

I can still see it being user error BUT it does seem a little prone to these issues for users.

So that's where the issue lies? Not where the 4 cables from the PSU meet the cable but where it plugs into the GPU?

It sounds like a user error thing... I mean... the cable clicked into position... there doesn't appear to be any stress on it although I do admit I'm not a fan of the design. The plug could have been underneath or on the side of the GPU... instead of the front where it is so close to the glass and is in the awkward positioning coming out and then down.

Edit: Oh wait.... just read the thread title again... so this was the old issue and now it's on the PSU side?

Wow! The gift that just keeps on giving! Either way... all my PSU cables are secured in the same fashion. Nothing I'm worried about.