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

[stahlhart]

Owning an Nvidia card has become a major issue. It's impossible to use it safely without 6-amp clamps and real-time monitoring or thermocouples.

Wow, drama queen much?

/ignore

 

[Bomby569]

But RT, DLSS and FG just work, who cares about cable, just replace it when it melts down.

remember: "the more you buy, the more you save"

 

[Dragam1337]

remember: "the more you buy, the more you save"

And just buy a 5070, it's faster than a 4090... according to jensen anyways, lol.

 

[Bomby569]

How is the circuit designed on the power supply unit, that the wires get evenly the amps? (watt = volt * amps)

let's not blame this on the PSU's, nothing failed on their side, they delivered the current that the gpu required, it's not the PSU's fault the card can't balance the load it requests.

 

[_roman_]

I just scrolled a bit back from romans video. he showed bad connectors on the psu side also.
Why would there be "burn marks" on the power supply side also?

I just searched the video again and made my fabulous picture. That thing looks like a power supply unit output connector from the high quality brand ASUS.

It's a combination of the graphic card, the user, the user settings in software known or unknown, the cable, the mainboard and the power supply unit. How many different psu types and brands are affected? Just ASUS with the ROG tax (republic of gamers tax)

 

[deathjester]

Looks like there was a melted cable issue. Most likely from a 3rd party cable. Just to be safe though i was wondering how many shunt resistors the Suprim SOC has on it? Buildzoid mentions that Asus has a bunch on theirs which is good. Can anyone tell me if MSI has a lot as well, or is it minimal. Basically wondering the likelyhood of my card burning up due to cheap manufacturing like NVIDIA.

Hey man, any information you can share about this?

If you look at the high res PCB shots from the review on here it looks like it only has 2 (which is double the founders edition), but it doesn't have the 6 in front of it that the ASUS cards have (again these can be seen on the rear of the PCB in the high res shots from the card review.
MSI 5090 Suprim

ASUS Astral (Front)

Rear

De8auer got his hands on it, and did some testing with his own 12VHPWR connector, and he was seeing 1 wire on the FE edition pulling 22 amps.
That said he has the 12VHPWR connector attached to 2 PCIE outputs from a Corsair AX1600i, but the official connector has this attached to 4.

Effectively he's trying to pull 300W over each of the PCIE connectors which is double what they are rated at, which would explain some of the issues; but not completely explain the fact that there's such an uneven current distribution across each of the 12V connectors; 23 Amps vs 11, 8, 5, 3 & 2 Amps.

Both of the issues I've seen have been with the FE model of the card and not with any of the AIBs, they are also with 12VHPWR ATX 3.0 PSUs rather than ATX 3.1 PSUs with the 12v-2x6 connector.

 

[Bomby569]

I just scrolled a bit back from romans video. he showed bad connectors on the psu side also.
Why would there be "burn marks" on the power supply side also?

I just searched the video again and made my fabulous picture. That thing looks like a power supply unit output connector from the high quality brand ASUS.


It's a combination of the graphic card, the user, the user settings in software known or unknown, the cable, the mainboard and the power supply unit. How many different psu types and brands are affected? Just ASUS with the ROG tax (republic of gamers tax)

that's just a messed up conclusion. If the cable was insanely hot it melted, and the cable was connected to both the gpu and psu so it melted itself and both ends, that can't take you to the conclusion the PSU is the problem, it's not the PSU's job to balance the load between the cables

edit: wires not cables

 

[TheDeeGee]

Anyways, Der8auers video now has proven that the FE has a design flaw, because it lacks per pin sensing, which the ASUS Astral 5090 does have and thus can balance the load over all six 12V cables.

I take it, back appears it ASUS can't balance the load. All it does is detect if one of the 6 wires has failed.

Either PSU's or GPU's need new circuitry for this, atleast one of the devices needs to know what's happening, that's for sure.

 

[_roman_]

This is your viewpoint. That is okay.

Let's assume there is over current protection per pin basis on that connector. Would that issue happen?


MY spare psu: https://www.chieftec.eu/products-detail/544/POLARIS_3.0_SERIES_/546/PPS-1050FC-A3
I took it as my personal reference to check

Security	AFC (Automatic Fan speed Control) OPP (Overload protection) OTP (Over Temperature Protection) OVP (Overvoltage protection) SCP (Short-circuit protection) SIP (Surge & Inrush Protection) UVP (Under Voltage protection) OCP (Over-Current Protection)

So a melted connector is totally acceptable for a ASUS psu? The device should be designed that regardless what I connect it will not melt the connector.

I do not have the schematics of the asus psu. Or the specs for the parts involved. Up to 87.5 Amps at 12V DC sounds very fun what my chieftec psu can deliver. I'm well aware of that is the up to max for all the 12V DC lines.

 

[sbacc]

I take it, back appears it ASUS can't balance the load. All it does is detect if one of the 6 wires has failed.

Either PSU's or GPU's need new circuitry for this, atleast one of the devices needs to know what's happening, that's for sure.

Ideally speaking, we need more security margin for current (for 600w we need 8 16AWG wires) and current load management and monitoring circuitry on both end (PSU & GPU). That means we need a new connector and a new standard for PSU as well...

Edit (why 8x 16 AWG, because we will still be OK and have acceptable margin when some companies decide to go cheap, cheat and send product with 18 AVG wires...)

 

[inquisitor1]

this is awesome. may help push some people to buy amd cards. im not going to 40 or 50 till its all settled and fixed finally. this nvidia connector is a blessing for amd.

 

[Bomby569]

This is your viewpoint. That is okay.

Let's assume there is over current protection per pin basis on that connector. Would that issue happen?


MY spare psu: https://www.chieftec.eu/products-detail/544/POLARIS_3.0_SERIES_/546/PPS-1050FC-A3
I took it as my personal reference to check

Security	AFC (Automatic Fan speed Control) OPP (Overload protection) OTP (Over Temperature Protection) OVP (Overvoltage protection) SCP (Short-circuit protection) SIP (Surge & Inrush Protection) UVP (Under Voltage protection) OCP (Over-Current Protection)

So a melted connector is totally acceptable for a ASUS psu? The device should be that designed that regardless what I connect it will not melt the connector.

I do not have the schematics of the asus psu. Or the specs for the parts in involved. Up to 87.5 Amps at 12V DC sounds very fun what my chieftec psu can deliver. I'm well aware of that is the up to max for all the 12V DC lines.

i don't think you want facts but to confirm what you think.
In the same video derbauer did a test with is own gpu and PSU, a corsair and the same happened. Did you missed that part?

 

[TheDeeGee]

Ideally speaking, we need more security margin for current (for 600w we need 8 16AWG wires) and current load management and monitoring circuitry on both end (PSU & GPU). That means we need a new connector and a new standard for PSU as well...

Yeah so, either GPU makers need to stop cheaping out, or PSU makers need to rework the 12V-2x6 circuitry.

Atleast one of the devices needs to know what's happening with each of the six 12V cables.

 

[qxp]

I just scrolled a bit back from romans video. he showed bad connectors on the psu side also.
Why would there be "burn marks" on the power supply side also?

I just searched the video again and made my fabulous picture. That thing looks like a power supply unit output connector from the high quality brand ASUS.
View attachment 384369

It's a combination of the graphic card, the user, the user settings in software known or unknown, the cable, the mainboard and the power supply unit. How many different psu types and brands are affected? Just ASUS with the ROG tax (republic of gamers tax)

Actually melting at both sides is what you would expect. With cables a large (or most) of the resistance comes from contacts. A little oxidation can increase it by a large factor. Because NVidia just combined wires some wires randomly can get a lot more current than they or the pins they are connected to were designed for. So now you have a lot of current going through a pin with some resistance - meltdown.

To see my point, let's assume that the pin contact resistance is only 0.01 Ohm (it can easily range from 0.01 Ohm to 0.1 Ohm). The power dissipated is current squared times resistance. So for 20A current you get (20A)^2*0.01Ohm = 4W. That's like a miniature light-bulb (the old type). As the contact heats up resistance increases, power dissipation goes up, more oxidation happens, more resistance, melting.

 

[Bomby569]

Ideally speaking, we need more security margin for current (for 600w we need 8 16AWG wires) and current load management and monitoring circuitry on both end (PSU & GPU). That means we need a new connector and a new standard for PSU as well...

wouldn't that means more cost on the psu side because nvidia cheap'd out on their hardware, we pay for their cheapness.
4000usd gpus and we need more expensive psus to protect their under engineered cards.

 

[Assimilator]

I also have a current clamp TRUE RMS multimeter.

Around minute 15

I wonder how many times this video is going to be posted in this thread...

 

[_roman_]

#184

In the same video derbauer did a test with is own gpu and PSU, a corsair and the same happened. Did you missed that part?

#173 i do not think so.

Is it not the power supply unit fault when one wire gets according to roman something around 10 Amps and the other wire only 2 amps?

#185 That guy finished my thought.

... PSU makers need to rework the 12V-2x6 circuitry.

Anyway Step 7 of the 8D-Report. What is permanently done to fix that issue?
A different coloured connector is not a fix.

-- roman had a good thought. Every non nvidia cable is a third party cable. This classifies a power supply cable from the power supply unit also as third party cable. It's just a connector

How comes it's always ASUS? Did I hit a nerve? Do not worry I have a ASUS Proart PA278QV monitor and a ASUS Prime x670-P Mainboard. Hardware I bought myself in past few years.

It is the psu task to deliver the voltage and amps correctly to the connectors. Not the mainboard, sata ssd, sata hdd, graphic card, cpu cooler, pc case, keyboard, mouse, usb audio interface task.
It's the cables and connectors task to be in spec. To work with different brands.

 

[sbacc]

wouldn't that means more cost on the psu side because nvidia cheap'd out on their hardware, we pay for their cheapness.

Having more control on load balancing could also improve efficiency and with good quality PSU you can keep those for a decade, paying a few bucks more for better power delivery and security is worth it in my opinion...

 

[oxrufiioxo]

I don't think anyone is denying there are also cables and GPUs working properly. The issue is the luck of the draw and the amount of things you get burdened with to pull luck in your favor.

Such as your inspection.

Yeah, not a single 4090 build I've put together has melted... That being said after watching the Buildzoid video the only 5090 I'd be comfortable owning is the Astral and I can't believe a company as technically capable as Nvidia doesn't have a failsafe if multiple pins have poor contact.

 

[LittleBro]

let's not blame this on the PSU's, nothing failed on their side, they delivered the current that the gpu required, it's not the PSU's fault the card can't balance the load it requests.

Exactly but still, I'd like a safety measures for that as well on PSU side.

I just scrolled a bit back from romans video. he showed bad connectors on the psu side also.
Why would there be "burn marks" on the power supply side also?

I just searched the video again and made my fabulous picture. That thing looks like a power supply unit output connector from the high quality brand ASUS.
View attachment 384369

It's a combination of the graphic card, the user, the user settings in software known or unknown, the cable, the mainboard and the power supply unit. How many different psu types and brands are affected? Just ASUS with the ROG tax (republic of gamers tax)

Melting on both ends of same wire is exactly what was supposed to happen. It clearly shows the GPU requested such high current through that pin, it got delivered and it also damaged the PSU. This also points to a fact that connector is a problem. Were the connector resistances spread evenly across connector, that wire would not get fed with 22A instead of 6A and would not heat up so much that it melts the wire isolation.

It's simply not like with old school ATX connectors anymore. Contact (mating) surface is much smaller with these 12 pin connectors and that paves road for higher resistance. In terms of wires, resistance is unwanted thing, a parasitic thing. It will always be there, you just need to design things in a way that compensates for it. Which this connector fails to do.

This is your viewpoint. That is okay.

Let's assume there is over current protection per pin basis on that connector. Would that issue happen?


MY spare psu: https://www.chieftec.eu/products-detail/544/POLARIS_3.0_SERIES_/546/PPS-1050FC-A3
I took it as my personal reference to check

Security	AFC (Automatic Fan speed Control) OPP (Overload protection) OTP (Over Temperature Protection) OVP (Overvoltage protection) SCP (Short-circuit protection) SIP (Surge & Inrush Protection) UVP (Under Voltage protection) OCP (Over-Current Protection)

So a melted connector is totally acceptable for a ASUS psu? The device should be designed that regardless what I connect it will not melt the connector.

I do not have the schematics of the asus psu. Or the specs for the parts involved. Up to 87.5 Amps at 12V DC sounds very fun what my chieftec psu can deliver. I'm well aware of that is the up to max for all the 12V DC lines.

These protections are not applied to every connector, rather to rails. It's user's duty to evenly distribute load. There is no per connector current sensing or fuse. Though I insist on PSU makers to start implementing it. I will gladly pay $30 more to have this stuff. For safety sake. This shit is dangerous.

 

[qxp]

I started a new thread with a proposal for new power connector - please comment:

Proposed new Power Connector

With all the attention on melting graphics card connectors, this might be an opportunity to switch manufacturers to a connector that we like. Here is my proposal - please comment whether you like the idea or not, any changes and if we get enough people behind it we'll try to make them listen...

www.techpowerup.com

 

[Bomby569]

Exactly but still, I'd like a safety measures for that as well on PSU side.

that's a fair ask, but that shouldn't be a thing so Nvidia can cheap out, it would make no sense. Because the PSU can always fail and we'd be moving responsability

 

[sbacc]

I understand the POV of users about why our PSU need to change just because of new Nvidia GPUs, but the market seems to want bigger and more power hungry GPU for all those new ""shiny"" feature and the market also want less cluter for those glass RGB showcase PC. For all that, a new cable was needed.
Personally I don't care for any of those (AMD user and mesh case), but if a new PSU standard (and a new cable/connector) can protect someone's home from burning, I am all for it.

(edited for more clarity and better context)

 

[Assimilator]

Honestly, the correct answer is for per-pin current sensing to be on both PSU and GPU side.

That said I will once again agree that not including current sensing and balancing on the FE is a pants-on-head-retarded move by NVIDIA. ASUS, for once, did something right.

 

[evernessince]

You inspecting it every couple of months and spraying it with deoxid says enough about the connector doesn't it?

Should be plug in and forget.

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.