It's happening again, melting 12v high pwr connectors

[deathjester]

Okay, well that's good to know. I just bought a new 3.1 ATX NZXT C1500 PSU.

Guess I'm wondering about that video from De8eaur and then Buildzoid. He basically says NVIDIA made the 12vhpwr even worse (saying he was detecing 150c when the card was running from the PSU side (which is unbelievable and that's on a test bench), and that there is nothing on the pcb to tell the card that the pins are not fully connected. Lots of comments on those vidoes and on reddit feel like we're going to be getting a lot more failures. Good to hear that the SUpric SOC has 2 shunt resistors, but now I'm worried about the card. I haven't received it yet, and I have 4090, and am wondering if it makes sense to return the 5090.

Also I'm using an right angled connect form cablemod, not the adapter (the right angles have had no problems reported), and am wondering if I should just switch to the connector that came with my PSU. I've had the card for 2 years and have had no issues, so likely the cable is fine, just wondering what you guys think. I also have an alert set up on HWINFO for if the 12vhpwr goes below 11.8v.

Thanks for the advice!

Thanks for the reply, please see my post above for additional concerns, wondering if you saw those videos and still feel that the issues have been resolved.

Yeah, I watched both of those.
there’s also a Spanish guy with the same issue:

but as I said both are ATX3.0 power supplies and FE cards.
However it is the same pin and its odd that De8auer is having similar issues, but again that’s over 2x8 pin PCIe where the adapter it comes with is 4x8 pin

 

[Bomby569]

They got away with it by deflecting that even it professionals are plugging them in wrong...

if you own the narrative and the narrators you can get away with murder

 

[Outback Bronze]

der8auer checked and found out that 1 cable goes up to 23 Amps

Yeah, did you see the heat coming off those lines too? Like 150°c from the back of the PSU. F me..

 

[Bones]

To me, if it's not right it's not right.
Doesn't matter who exposes it because if the problem is real, then it's a good thing it got called out.

I'll wait and see what "Solutions" appear to the issue.

 

[TheDeeGee]

To me, if it's not right it's not right.
Doesn't matter who exposes it because if the problem is real, then it's a good thing it got called out.

I'll wait and see what "Solutions" appear to the issue.

The only solution if they want to keep 12V-2x6 is proper circuitry which monitors and balances the six 12 volt cables, there is no other solution for this connector. This would also mean the 5090 could potentially get power limited, because one or more of the 12 volts cables arn't making a proper connection and the other wires would have a limit of how many Amps can flow through.

And this circuitry has to either be implented into PSU's, which also means 8-Pin connectors need this (because 2x 8-Pin to 12V-2x6 cables) and/or the GPU's.

That is the only proper protection.

 

[Space Lynx]

If people just use the new 2x6 cable that replaces the 12vphr or w.e it's called, I mean all of these issues could be avoided.

 

[TheDeeGee]

If people just use the new 2x6 cable that replaces the 12vphr or w.e it's called, I mean all of these issues could be avoided.

You have no idea what's actually the problem here do you?

 

[qxp]

If people just use the new 2x6 cable that replaces the 12vphr or w.e it's called, I mean all of these issues could be avoided.

If you look at the earlier comments, it looks like incompletely inserted connector was only a secondary cause. The real problem is design of the cards that has no means to load-balance current on the wires - they are tied together at the GPU and the power supply. So a slightly increased resistance at some of the contacts will concentrate current onto the other wires causing meltdown.

An improperly inserted connector is more likely to have different resistances at different pins, but this can happen anyway with a fully inserted connector too.

This is clearly bad design - small variation in contact or wire resistance should not affect distribution of current on the wires.

 

[dragontamer5788]

The cause of uneven draw isn't the GPU nor the PSU... they're just sending and receiving the power. In my simple mind that just leaves the cable being insufficient, the size inviting too much variance.

Just an FYI: a 4AWG wire aka 5mm thick is required to sustain a 50Amp / 600W load at 12Vs. Or 20mm^2 area, roughly. That'd be two wires (50Amps from power, and then 50 Amps to ground). The 12pin connector uses somewhere like 16AWG wire, which is only 1.3mm thick or ~1.3mm^2 area. Even adding all the area up for 6x lines (6 power and 6 ground == 12 total wires), that's only 8mm^2 best case scenario.

Yeah, bigger wire is better. But its clear that a full size lol 5mm thick / 4AWG gauge wire is too bulky for these applications.

The only way this works is if the power sent down the lines had some kind of balancing. PSUs traditionally never balanced power. That's always been the job of the cards to pull from each connector in a balanced manner. (If the 3x8 pin cards all pulled power from just one connector, no one would be surprised if they caught on fire. That'd be obviously a GPU Card's mistake). Similarly, if the 12-pin connector requires balancing between the lines to prevent these 150C / 300F temperature wires, then so be it. Its the card's responsibility to pull the appropriate amount of power and balance.

Isn't it also far more economical to just upscale the cable a good bit so you can handle the unbalanced load (and likely get less of it as there is more surface area) instead of desiring even more hardware on a GPU where the desire is to reduce PCB sizes?

Oh yeah. Everyone is wondering why they didn't just stick with 3x 8-pin or even 4x 8-pin connectors now. Its seemingly like this 12-pin issue was a epic fail entirely. Maybe its a little bit larger but you know? Maybe trying to shrink down a 50Amp connection is itself a mistake.

Yeah, did you see the heat coming off those lines too? Like 150°c from the back of the PSU. F me..

Heat = I^2 * R

10-Amps is 100x hotter than 1-Amp.

22-Amps is 480% hotter than 10-Amps.

If people just use the new 2x6 cable that replaces the 12vphr or w.e it's called, I mean all of these issues could be avoided.

By specifications, you need 4x8 pin cables to replace the 12-pin connector.

 

[x4it3n]

You have no idea what's actually the problem here do you?

I guess he didn't watch der8auer nor buildzoid videos lol. Nvidia are really cheaping out on everything this generation...

 

[evernessince]

If people just use the new 2x6 cable that replaces the 12vphr or w.e it's called, I mean all of these issues could be avoided.

Please read the thread before commenting. This had already been addressed in two videos and over many comments.

The short of it is no, the cable doesn't matter. They are identical anyways and it doesn't fix the fact that the current balancing capabilities across all the cable wires were basically neutered for the 4000 series and later. As buildzoid has pointed out, you can put all 600w of power through a single wire on a 12VHPWR / 12V2X6 equipped cable and the card won't complain, it will just melt.

Nvidia has never stated that the two are incompatible but as a hypothetical, let's assume that 12V2X6 aren't compatible. Ok, so Nvidia failed at basic engineering so hard they didn't key the connectors different. So in your purported version of events, Nvidia would still be to blame.

Of course that isn't the case but no matter which way you look at it, Nvidia screwed up.

 

[desiprofessor]

Okay so as I wrote, I've got the new NXZT Atx 3.1 C1500 PSU for the pre ordered MSI 5090 SOC. I also have a Suprim X 4090 right now, but that is connected to a Asus ATX 3.0 compatible Thor 1200Wp2.

The cable being used is a cablemod 90 degree cable (not the adaptor that was recalled) that splits into 3 8 pins into the PSU. I've used the 4090 for 2 years and have had no issues at all.

Do people think I shoudl cancel the 5090 and just stick with the 4090? Also if I were to stick with the 4090 should I switching the PSU cables original cables for the ASUS Thor, or even better switching it up to the the NZXT Atx 3.1 that I have in a box?

Suggestions woudl be greatly appreciated.

 

[deathjester]

If people just use the new 2x6 cable that replaces the 12vphr or w.e it's called, I mean all of these issues could be avoided.

It's not the cable that's 12v-2x6 it's the socket on the PSU / GPU that use that connector. The 12VHPWR cable doesn't change.

I guess he didn't watch der8auer nor buildzoid videos lol. Nvidia are really cheaping out on everything this generation...

Actually the Asus LOKI PSU and the one from the Spanish video above are both ATX 3.0 so has the 12VHPWR connector rather than the 12V - 2x6 pin, so he's not entirely incorrect.

De8auer is using 2x8 pin PCIe which are meant to be rated @ 150W rather than the 300W that he is trying to pull through them.
The bundled adaptor has 4x8 pin PCIe. Not that it should be pulling 23 Amps through one wire tho.

 

[TheDeeGee]

Okay so as I wrote, I've got the new NXZT Atx 3.1 C1500 PSU for the pre ordered MSI 5090 SOC. I also have a Suprim X 4090 right now, but that is connected to a Asus ATX 3.0 compatible Thor 1200Wp2.

The cable being used is a cablemod 90 degree cable that splits into 3 8 pins into the PSU. I've used the 4090 for 2 years and have had no issues at all.

Do people think I shoudl cancel the 5090 and just stick with the 4090? Also if I were to stick with the 4090 should I switching the PSU cables original cables for the ASUS Thor, or even better switching it up to the the NZXT Atx 3.1 that I have in a box?

Suggestions woudl be greatly appreciated.

Cancel the 5090.

Cable used for 4090 doesn't matter as it has the exact same issue as the 5090.

If anything, a 2x 8-Pin to 12V-2x6 cable could be considered the safest, as the PSU end has thicker pins then.

 

[desiprofessor]

Cancel the 5090.

Cable used for 4090 doesn't matter as it has the exact same issue as the 5090.

If anything, a 2x 8-Pin to 12V-2x6 cable could be considered the safest, as the PSU end has thicker pins then.

So can't I theoretically do that with the 5090? Just use the NVIDIA adaptor and go 4 8 pins into the psu?

 

[Vayra86]

the problem is it doesn't make sense, it isn't the cable that decides were the load goes, it's just a cable.

Its a cable with connectors on it that have variances in how they are built, how well they connect and as a result how much resistance they have

Its also a cable that doesn't handle those variances well because there the whole thing is too flimsy

 

[evernessince]

Okay so as I wrote, I've got the new NXZT Atx 3.1 C1500 PSU for the pre ordered MSI 5090 SOC. I also have a Suprim X 4090 right now, but that is connected to a Asus ATX 3.0 compatible Thor 1200Wp2.

The cable being used is a cablemod 90 degree cable that splits into 3 8 pins into the PSU. I've used the 4090 for 2 years and have had no issues at all.

Do people think I shoudl cancel the 5090 and just stick with the 4090? Also if I were to stick with the 4090 should I switching the PSU cables original cables for the ASUS Thor, or even better switching it up to the the NZXT Atx 3.1 that I have in a box?

Suggestions woudl be greatly appreciated.

Honestly I wouldn't change anything. With the way the connector is wired now, changing things out is a gamble. If you happen to get a cable with a defect or a card with a less than ideal receptacle, you won't know until your card is permanently damaged. It will happily run all the power through a single wire until it melts. There are no mechanisms in place to prevent the card from running with a bad cable unless every wire is bad, which is completely un-ideal.

This issue isn't just with the 5090 as well. The 4090 and potentially even the 4080 could be problematic. After all, the 4080 still has a TDP north of 300w and 300w in a worst case scenario through a single wire is still not good.

There should be a recall of 5090 and 4090 at the very least IMO. There should have been a recall of the 4090 to begin with when this issue first started cropping up. I power limit my 4090 to 65% and have never removed the cable but even then after the latest info from buildzoid I'm pretty ticked that such a defective design was not recalled. Instead Nvidia just doubled down on passing the blame.

 

[desiprofessor]

Buildzoid farming clicks with Drama again?
i supported him on patreon for almost half a year but now i click "do not show any videos from this channel" when i see him on YT.
ONE Case of a space constrained SFF PC with a old third party cable? oh no!

Did you see De8auer's video, he is having the same issues with his own GPU.

Honestly I wouldn't change anything. With the way the connector is wired now, changing things out is a gamble. If you happen to get a cable with a defect or a card with a less than ideal receptacle, you won't know until your card is permanently damaged. It will happily run all the power through a single wire until it melts. There are no mechanisms in place to prevent the card from running with a bad cable unless every wire is bad, which is completely un-ideal.

This issue isn't just with the 5090 as well. The 4090 and potentially even the 4080 could be problematic. After all, the 4080 still has a TDP north of 300w and 300w in a worst case scenario through a single wire is still not good.

There should be a recall of 5090 and 4090 at the very least IMO. There should have been a recall of the 4090 to begin with when this issue first started cropping up. I power limit my 4090 to 65% and have never removed the cable but even then after the latest info from buildzoid I'm pretty ticked that such a defective design was not recalled. Instead Nvidia just doubled down on passing the blame.

So you think i should stick with the 5090?

Should I use the cable that comes with the GPU, it's an adaptor that has space for 4 8 pins into the PSU. Or should I go with the 12v 2x6 cable that is the same on both sides that goes into my NZXT ATX 3.1 PSU?

Thanks!

 

[jnv11]

It's not the cable that's 12v-2x6 it's the socket on the PSU / GPU that use that connector. The 12VHPWR cable doesn't change.


Actually the Asus LOKI PSU and the one from the Spanish video above are both ATX 3.0 so has the 12VHPWR connector rather than the 12V - 2x6 pin, so he's not entirely incorrect.

De8auer is using 2x8 pin PCIe which are meant to be rated @ 150W rather than the 300W that he is trying to pull through them.
The bundled adaptor has 4x8 pin PCIe. Not that it should be pulling 23 Amps through one wire tho.

der8auer is using an adapter cable with 2 8-pin Corsair type 4 plugs on the PSU side, not 2 8-pin PCI Express Graphics (or PEG) power plugs. Those are two different standards.

The 6-pin and 8-pin PCI Express Graphics power connector standards were constrained by the fact that when they were defined, they needed to be adapted from Molex connectors, and many Molex connectors were fed by thin gauge wires, poor quality pins in the Molex plugs, and poor quality sockets in the Molex sockets. Furthermore, the people defining those sockets knew that PSU manufacturers would cheap out on the wires, sockets, and plugs back then. Also, a 6-pin PEG connector only has 2 defined 12V pins (a third that is frequently powered is officially a no connection) and 3 ground pins, while an 8-pin PEG connector has 3 defined 12V pins and 5 ground pins. In both of these, one of the grounds can be a sense pin. These reasons are why 6-pin PEG sockets are specified up to 75 watts maximum, and 8-pin PEG sockets are specified up to 150 watts maximum.

Corsair decided to focus on the high end when defining its proprietary type 4 connectors, so it requires thick gauge wires, good quality plugs, and good quality sockets, allowing each wire to carry far more amps safely as long as the device on the other side does not wildly misbehave with its power draw. Furthermore, Corsair type 4 8-pin connectors have 4 12V pins and 4 ground pins.

der8auer is safe in using 2 8-pin Corsair type 4 plugs on the PSU side of the adapter cable. Your asertion that his use of 2x8 pin PCIe connectors is wrong due to misidentifying which standard they belong to.

 

[evernessince]

Did you see De8auer's video, he is having the same issues with his own GPU.



So you think i should stick with the 5090?

Should I use the cable that comes with the GPU, it's an adaptor that has space for 4 8 pins into the PSU. Or should I go with the 12v 2x6 cable that is the same on both sides that goes into my NZXT ATX 3.1 PSU?

Thanks!

Actually how long is your pre-order out from now? I was just thinking if it's far enough out it might be wise to wait to see if anything comes out that can monitor the cable. I know wireview can monitor partial connections but perhaps it will be upgraded to ensure per pin power draw.

Otherwise I think the adapters that convert from 8 pin are best or if your PSU has 8-pin on the PSU side.

 

[JAS_design]

I apologise if this has been mentioned before; I have read the thread but may have missed something.

What I've noticed is that most of the melting seems to be on the positive pins, rather than the negative pins. I think that could be important.

It's my understanding that the 12vhpwr connector is a customised version of the Amphenol Minitek Pwr 3.0. I've never used those, but I have used the very similar Molex MicroFit 3.0 in my personal projects for many years, and can offer some of my personal observations.

Firstly, the MicroFit 3.0 does require more care and a little more force to fully insert than the larger MiniFit Jr. I've never had a problem with it, but I can see how other people might.

Perhaps more importantly, Molex make it very clear that MicroFit 3.0 PCB Headers should be soldered to a large copper plane, to ensure proper cooling of the soldered pins. I haven't seen this mentioned in the documentation for the Amphenol Minitek connectors, but it might explain why only positive pins are melting. Negative pins are usually soldered to a large ground plane, but positive pins can sometime be connected to comparatively much smaller traces.

If the design of the PCB is providing less cooling for the positive pins, that might also be part of the problem.

 

[N/A]

Corsair decided to focus on the high end when defining its proprietary type 4 connectors, so it requires thick gauge wires, good quality plugs, and good quality sockets, allowing each wire to carry far more amps safely as long as the device on the other side does not wildly misbehave with its power draw. Furthermore, Corsair type 4 8-pin connectors have 4 12V pins and 4 ground pins.

der8auer is safe in using 2 8-pin Corsair type 4 plugs on the PSU side of the adapter cable. Your asertion that his use of 2x8 pin PCIe connectors is wrong due to misidentifying which standard they belong to.

3x12V per typ 4 goes to 3x12V on the GPU side, same difference, and 1GND goes for sense pins. Only 8 pin EPS gets 4 GND and 4x12V

 

[jnv11]

3x12V per typ 4 goes to 3x12V on the GPU side, same difference, and 1GND goes for sense pins. Only 8 pin EPS gets 4 GND and 4x12V

Wrong. See https://help.corsair.com/hc/en-us/article_attachments/17759705845389 to see that a Corsair type 4 8-pin connector has 4 12V pins and 4 ground pins.

 

[Rahmat Sofyan]

it was start with rtx 3000 series, nvidia prefer clean and compact devices ... but overall how this cards passed ce eu or other requirements ?

 

[Visible Noise]

Yep, makes you wonder how bad this situation would be if people didn't take so many steps to try and prevent it already.

It's paradoxical, you have all these people claiming the connector is fine yet you are an idiot if you don't follow what seems like a growing 12VHPWR / 12V2X6 guide book on how not to destroy your GPU. Now you can't even use 12VHPWR cables or any third party cables without being it immediately being called "user error". It contradicts the reason to release a public spec in the first place.

And yet in 14 months I haven’t bothered to take a single glance at my 4090‘s power cable.

Maybe I’m not an idiot that doesn’t know how to properly seat a plug I guess?

Stumbled on this x thing.View attachment 384384

So we’re back to user error.

Who’d a thunk?