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

[Dredi]

but like you say... should come with the connectors on the gpu then.

Or in this case, it could be done on a pin per pin basis, like in the 3090ti (with like almost zero melted cable issues). Ofc. the thermal limit is a bit too close to comfort on a single 12pin and having more heat dissipation area, by using more connectors, would be a good idea in case someone uses an out of spec cable or fails in it’s installation.

 

[Degreco]

Building a computer is not idiot proof. You need to be handy, have specific knowledge and skill.

Nah. Nowadays it's very easy compared to the old days.

 

[AVATARAT]

Building a computer is not idiot proof. You need to be handy, have specific knowledge and skill.

Are you saying that people need a technical education to plug in the cables specifically on the Nvidia cards?
Are you saying that when you buy a card for a few thousand dollars you should buy better cables for it because the ones Nvidia provides are cheap junk?

Just asking

 

[BoggledBeagle]

Are you saying that people need a technical education to plug in the cables specifically on the Nvidia cards?
Are you saying that when you buy a card for a few thousand dollars you should buy better cables for it because the ones Nvidia provides are cheap junk?

Just asking

No, that is not what I am saying.

Regarding inserting the plug you should in my opinion follow these general guidelines:

1. Make sure that the cable you are using is in good condition, especially the plugs. In case of brand new cables from reputable producers you may skip this step, but of course you could look in the plugs to see, if any foreign object did not fall in the plugs during packing and shipping.
2. In case the wires are bent from the packing of the cable, straighten the cable fully.
3. According to the planned location of the wires, bend or form the wires in advance to make mounting the wires on the chassis and inserting the plug easier.
4. Support the object into which you insert in the moment of insertion, so that the object does not deform, or move, to avoid any damage.
5. Insert the plug straight and fully, make sure that the latching mechanism secures the plug in place.
6. Mount the cable leading to the plug on the computer chassis so that there is a slack of wire at the plug and the wire does not exert any unnecessary force on the plug.
7. Make sure that after closing the PC case or placing other components in the case the condition in point 6 does not change

Any ideas for other recommendations?

 

[AVATARAT]

No, that is not what I am saying.

Regarding inserting the plug you should in my opinion follow these general guidelines:

1. Make sure that the cable you are using is in good condition, especially the plugs. In case of brand new cables from reputable producers you may skip this step, but of course you could look in the plugs to see, if any foreign object did not fall in the plugs during packing and shipping.
2. Support the object into which you insert in the moment of insertion, so that the object does not deform, or move, to avoid any damage.
3. Insert the plug straight and fully, make sure that the latching mechanism secures the plug in place.
4. Mount the cable leading to the plug on the computer chassis so that there is a slack of wire at the plug and the wire does not exert any unnecessary force on the plug.
5. Make sure that after closing the PC case or placing other components in the case the condition in point 4 does not change

Any ideas for other recommendations?

Hum, sounds interesting, if someone gets a Radeon card they'll just plug it in and play, but if they get a 16-pin Nvidia card they have to follow the guidelines. I hope you understand the difference.

See, we all know these steps, and we know we have to be very careful with this 16-pin crap, and we're commenting on just how poorly made it is.

 

[ShrimpBrime]

No, that is not what I am saying.

Regarding inserting the plug you should in my opinion follow these general guidelines:

1. Make sure that the cable you are using is in good condition, especially the plugs. In case of brand new cables from reputable producers you may skip this step, but of course you could look in the plugs to see, if any foreign object did not fall in the plugs during packing and shipping.
2. In case the wires are bent from the packing of the cable, straighten the cable fully.
3. According to the planned location of the wires, bend or form the wires in advance to make mounting the wires on the chassis and inserting the plug easier.
4. Support the object into which you insert in the moment of insertion, so that the object does not deform, or move, to avoid any damage.
5. Insert the plug straight and fully, make sure that the latching mechanism secures the plug in place.
6. Mount the cable leading to the plug on the computer chassis so that there is a slack of wire at the plug and the wire does not exert any unnecessary force on the plug.
7. Make sure that after closing the PC case or placing other components in the case the condition in point 6 does not change

Any ideas for other recommendations?

Just stick with the 12 male female pins are too small and there's too much resistance in the connector to pull that much current.

Typically, when you want higher amp draw (power), you use a larger cable and a larger connector.

It's like trying to start a semi-truck with 2awg cable but a starter capable of 200a pull on 12v turning a 15 Liter motor. The cable size should be 2 aught... for example.

 

[Dredi]

No, that is not what I am saying.

Regarding inserting the plug you should in my opinion follow these general guidelines:

1. Make sure that the cable you are using is in good condition, especially the plugs. In case of brand new cables from reputable producers you may skip this step, but of course you could look in the plugs to see, if any foreign object did not fall in the plugs during packing and shipping.
2. In case the wires are bent from the packing of the cable, straighten the cable fully.
3. According to the planned location of the wires, bend or form the wires in advance to make mounting the wires on the chassis and inserting the plug easier.
4. Support the object into which you insert in the moment of insertion, so that the object does not deform, or move, to avoid any damage.
5. Insert the plug straight and fully, make sure that the latching mechanism secures the plug in place.
6. Mount the cable leading to the plug on the computer chassis so that there is a slack of wire at the plug and the wire does not exert any unnecessary force on the plug.
7. Make sure that after closing the PC case or placing other components in the case the condition in point 6 does not change

Any ideas for other recommendations?

Is this the exhaustive list of documented steps one should have for you to deem that a burnt connector was good and correctly installed?

 

[Jacky_BEL]

How?

Use with damaged cables, incorrectly installed or incorrectly manufactured cables is irrelevant.

Provide a link to a single case of a correctly installed, undamaged and well made cable that caused a problem. It should be well documented and reliable.

Is Derbauer's video about unbalanced load in the cables enough? Go check it out.

 

[Endymio]

Are you saying that when you buy a card for a few thousand dollars you should buy better cables for it because the ones Nvidia provides are cheap junk?

Just asking

NVidia doesn't provide cables. Any more dumb questions?

Is Derbauer's video about unbalanced load in the cables enough? Go check it out.

No, it's not enough, especially when he outright admits that (a) his statements refer to the one specific cable he tested, not all cables in general, and (b), also admits that, "essentially no one else could replicate" his results.

 

[ShrimpBrime]

NVidia doesn't provide cables. Any more dumb questions?

Nvidia provides the adapter cable when you purchase the NV FE cards.

My 4070S from MSI came with the adapter, but I feel safe pulling only 220w on it. I feel this card would be ok on a single 8 pin. (This is NV trying to standardize garbage most of their cards don't actually need.)

 

[Endymio]

Nvidia provides the adapter cable when you purchase the NV FE cards.

And those aren't failing. The only case I know involving a FE 5090:

"GeForce RTX 5090 Founders Edition card suffers melted connector after user uses third-party cable...."

And in that case, the cable melted on the PSU end as well....

 

[Dredi]

The only case I know involving a FE 5090:

There are tens of them failing. Some with psu supplied cables from seasonic.

 

[BoggledBeagle]

There are tens of them failing. Some with psu supplied cables from seasonic.

You can destroy any cables. As I wrote in the Survey thread and here already too, a cable supplied with my Seasonic PSU had a plug made of soft plastic and that alone would make it less resistant to incorrect installation.

 

[Dredi]

You can destroy any cables. As I wrote in the Survey thread and here already too, a cable supplied with my Seasonic PSU had a plug made of soft plastic and that alone would make it less resistant to incorrect installation.

Did you test if it was within spec or no?
What kind of documentation did you produce? I call bullshit unless enough documentation is given.

 

[Endymio]

There are tens of them failing. Some with psu supplied cables from seasonic.

Did you not read my post? A Seasonic cable is not an NVidia cable.

I'll make it simpler. An engineer designs a foot bridge capable of supporting 650 amps...err, tons. A contractor builds several of these, but skimps on the materials and assembles them improperly. Some of the bridges last years, because they're only being loaded to 200 - 300 tons. Then one day a bridge gets loaded to 600 tons and collapses. Who's to blame -- the engineer, or the contractor?

These complaints are even more absurd when one realizes the engineer here isn't NVidia, but PCI-SIG.

 

[ShrimpBrime]

And those aren't failing. The only case I know involving a FE 5090:

"GeForce RTX 5090 Founders Edition card suffers melted connector after user uses third-party cable...."

And in that case, the cable melted on the PSU end as well....

Not yet? Could happen. The "what if's" about it.

If one plastic melts and the other looks burnt, then the molding is different.

My point here is the little deutsch connector is too small, and as I had proposed a few pages back, keep the design, lower the resistance with gold plating, which is cheap, as a resolve.

Otherwise, larger wires/cables and larger connectors.

The first problem is the male pin diameter looks to be smaller than the wire diameter. (Surface area. Tin over nickel plating. (Or was it nickle over tin, fk can't remember. Lol)

The second problem is the tiny female side lacks the grip of larger female connectors.

And when it comes to electricity, I always round UP. That's 10a per pin being pulled across it. You have to calculate the losses when factoring the transformation of 12v to 1.1-1.2v.

Maybe the NVidia provided one is just a tiny bit better where it doesn't melt.

The point is, why even try to melt it?
The original failures where feom people just gaming trying to enjoy the hardware.

NV dropped the ball.
If they want to continue 600w+ production, standardize PSUs with larger wire size for PCIE and utilize larger male female connectors on the card/cable and there will be less problems.

 

[BoggledBeagle]

Did you test if it was within spec or no?
What kind of documentation did you produce? I call bullshit unless enough documentation is given.

That is a good approach, first we must wait for the well documented and reliable case that a correctly installed, undamaged and well made cable caused a problem.

 

[Dredi]

Did you not read my post? A Seasonic cable is not an NVidia cable.

Are there any first party cables then? No.

That is a good approach, first we must wait for the well documented and reliable case that a correctly installed, undamaged and well made cable caused a problem.

Just commenting on the idiocity that at the same time you demand documentation, but are incapable of defining said documentation.

 

[Maxiaid]

Regarding inserting the plug you should in my opinion follow these general guidelines:

1. Make sure that the cable you are using is in good condition, especially the plugs. In case of brand new cables from reputable producers you may skip this step, but of course you could look in the plugs to see, if any foreign object did not fall in the plugs during packing and shipping.
2. In case the wires are bent from the packing of the cable, straighten the cable fully.
3. According to the planned location of the wires, bend or form the wires in advance to make mounting the wires on the chassis and inserting the plug easier.
4. Support the object into which you insert in the moment of insertion, so that the object does not deform, or move, to avoid any damage.
5. Insert the plug straight and fully, make sure that the latching mechanism secures the plug in place.
6. Mount the cable leading to the plug on the computer chassis so that there is a slack of wire at the plug and the wire does not exert any unnecessary force on the plug.
7. Make sure that after closing the PC case or placing other components in the case the condition in point 6 does not change

Any ideas for other recommendations?

That's the most painfully obvious guide I've ever read. And you said you need to be handy, have specific knowledge and skill for that?

 

[Endymio]

Not yet? Could happen. The "what if's" about it.

And the sun could fail to rise tomorrow as well. Thanks for admitting no NVidia products are failing.

My point here is the little deutsch connector is too small....

Connectors that meet spec aren't failing. And even if your point were correct, the issue then lies with PCI-SIG, a consortium of 800+ companies, currently led by -- wait for it -- an IBM engineer.

, and as I had proposed a few pages back, keep the design, lower the resistance with gold plating, which is cheap, as a resolve.

What do you think that would accomplish? Gold is worse than copper. It's better than aluminum ... but plating a pin with gold doesn't lower the resistance of the underlying aluminum underneath. Gold plating is only helpful if these failures were due to corrosion.

Are there any first party cables then? No.

I'm not sure why you're struggling with the logic here. NVidia didn't design the cable. It didn't build these substandard cables that are failing. Put down the pitchfork.

 

[Jacky_BEL]

NVidia doesn't provide cables. Any more dumb questions?


No, it's not enough, especially when he outright admits that (a) his statements refer to the one specific cable he tested, not all cables in general, and (b), also admits that, "essentially no one else could replicate" his results.

The problem is not with the cable , it is with the card not monitoring and exceeding current draw through the wires of the cables.

 

[Dredi]

I'm not sure why you're struggling with the logic here. NVidia didn't design the cable. It didn't build these substandard cables that are failing. Put down the pitchfork.

The cable spec demands that the current in one lead should not exceed X. Nvidia has failed to produce a gpu that would guarantee that the current per lead will not exceed X.

The cable CANNOT dictate the current pulled through it.

The PSU CANNOT dictate the current transferred through one lead.

 

[ShrimpBrime]

And the sun could fail to rise tomorrow as well. Thanks for admitting no NVidia products are failing.


Connectors that meet spec aren't failing. And even if your point were correct, the issue then lies with PCI-SIG, a consortium of 800+ companies, currently led by -- wait for it -- an IBM engineer.


What do you think that would accomplish? Gold is worse than copper. It's better than aluminum ... but plating a pin with gold doesn't lower the resistance of the underlying aluminum underneath. Gold plating is only helpful if these failures were due to corrosion.


I'm not sure why you're struggling with the logic here. NVidia didn't design the cable. It didn't build these substandard cables that are failing. Put down the pitchfork.

I mean, you make good points, but don't go out of your way to troll me.

If the cards where out 6 months, and a connector melted, maybe it would be a different story.

I can tell you, my 600w audio amplifier utilizes better connectors and heavier wire gauge and have never had a warm wire or connector, let alone a hot one to melt plastic. Same for the 12v supply.

 

[JustBenching]

Nobody ever demonstrated that a new correctly installed cable (or used cable with undamaged plugs) causes any problems.

NOBODY. EVER.

Yet this thread and bizzare comments still for some reason goes on.

I myself tested two new and like new cables with 400W load and they performed perfectly.
You are free to report your correct and factual information including how much have the plugs been used here:

12 pin GPU connector survey

I have a suspition, that not all cables and connectors are made with the care they deserve and there are some substandard and even dangerous cables out there. Some cables might have been damaged by users too. This survey should help to identify them. Post your photos of 12 pin GPU connectors...

www.techpowerup.com

Well obviously an undamaged connector won't cause issues, the question is how easy it is (if it is) to damage and how easily (if that's the case) it melts when damaged.

 

[lilhasselhoffer]

Did you not read my post? A Seasonic cable is not an NVidia cable.

I'll make it simpler. An engineer designs a foot bridge capable of supporting 650 amps...err, tons. A contractor builds several of these, but skimps on the materials and assembles them improperly. Some of the bridges last years, because they're only being loaded to 200 - 300 tons. Then one day a bridge gets loaded to 600 tons and collapses. Who's to blame -- the engineer, or the contractor?

These complaints are even more absurd when one realizes the engineer here isn't NVidia, but PCI-SIG.

You make the assumption that this is the case. Note, proof not given.

Now let me fix that example, and make it accurate.

You design a bridge to carry an average maximum of 600 tons. Each lane is specified to have a maximum load, which when combined add to your average maximum weight. You then have a new class of vehicle that travels at only 90 mph, but weighs 900 tons, and while traveling it's the only vehicle allowed on the bridge. Each lane on the bridge is assumed to take the same average weight, but you post-facto know that some lanes are basically loaded with semi-trucks that each weigh 500 tons...and the average loading is specced at and assumes 100 tons a lane. The bridge fails rather quickly, shearing itself from its moorings. Do you blame the engineer who designed the bridge, the cars allowed to transit, or maybe do a real investigation that shows those damnable 900 ton trucks combined with the 500 ton truck actually caused the bridge to fail in fatigue loading because the bridge was designed stupid, based not upon usage conditions but on a theoretical usage scenario which does not match reality?
Now, let's look at your 600 ton bridge. You haven't inputted a safety factor, and because you don't monitor anything those pesky 500 ton trucks and 900 ton trucks do damage. That's difficult to catch unless you see the trucks in motion...but all of this came about because you were told to minimize the cost of the bridge, and in doing so you removed the weight sensor designed to stop overloading...because that is why Nvidia introduced the new power connector.

Remember also that instead of over specifying the connection, Nvidia is pulling very nearly the maximum specification that can be delivered. If you want to be less obtuse, and you live in the Twin Cities area of Minneapolis and St. Paul (in Minnesota), then you'll remember the I-35 bridge collapse that was caused by this exact type of failure.


You'll also note that in sudden and catastrophic failures...like bridge collapse or bursting into flame...the blame was not immediately assigned to the user. It was tested, the specifications were verified, the actual conditions were found, and the resolution was that forcing users to accept the blame was an idiotic conclusion. Technically if the bridge had been limited to its design assumptions everything would have been fine...but those assumptions were made decades earlier with no real knowledge of what would be experienced in the real world, under the auspices of basically whatever the lowest bid was would win, and thus specifications were assumed met but not exceeded.