News 16-Pin Connectors Are Still Melting On RTX 4090 GPUs

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
As I said before (IMO) ... it is time to redesign the the power of the Motherboard GPU slot ... now that we are using Gen5 of PCIe .. and soon Gen 6 , USE 8 lanes only and the rest of the long x16 slot can be used for power delivery .. and cables be History !
 
I found the idea of a connector not fully seated causing a meltdown a weird explanation to begin with. I have never seen this happen with other connections.

Either there's a connection or there's not, there's no middle of the road where there's half a connection and the resistance somehow causes heat to build up, that seems like a fantasy explanation to blame the users.
In IT we tend to ignore just how unnatural binary is in real life: it's an abstract ideal, a reality that takes hard work to achieve in order to attain predictable and repeatable results, again difficult goals.

Connected or not, user error or not, liable or not, the answers won't be naturally binary.

What needs to be binary is the vendor answer and one of the things Intel did really right is the response on their FDIV bug: taking full responsibility and guaranteeing life-time free replacements.

Anything less (hear, hear AMDgeddon) will be far too expensive for the vendor in the long run.
 
Last edited:
Actually you CAN "idiot proof" power plugs . but it will cost you more. and thats the only reason.
And then an idiot will walk in and cram the cable in backward causing a melted cable. I have seen people force 8-pin cables in the wrong way. I do not understand how you believe you can mitigate all idiots. There are, like I and another have said, 50 or so cases world wide. I gave an example with 100 to be fair to those who have not reported some issues. If your design can survive tens of thousands, if not hundreds of thousands, of people and only have 50-100 idiots cause one hyper specific issue, you did a great job.
 
I found the idea of a connector not fully seated causing a meltdown a weird explanation to begin with. I have never seen this happen with other connections.

Either there's a connection or there's not, there's no middle of the road where there's half a connection and the resistance somehow causes heat to build up, that seems like a fantasy explanation to blame the users.
There is a middle ground: how much contact do the conductors have with each other. A smaller contact surface represents a thinner wire. A thinner wire has more resistance and carries less amps safely.
 
Gigantic cards in too small of a case is another huge factor here. Could be as simple as constant stress pushing on the power cable which in turn pushes on the GPU pcb (which probably has a lot of sag too) causing a heat build up in the card that leads to softening the plastic of the connector and leads to the wires/pins moving under the stress. With the heat damage covering up any evidence.

Most cards prior to the 30 series had ample room for the 8-pin connectors to not hit the side panels of cases, so a lot less instances of melted connectors.
 
And then an idiot will walk in and cram the cable in backward causing a melted cable. I have seen people force 8-pin cables in the wrong way. I do not understand how you believe you can mitigate all idiots. There are, like I and another have said, 50 or so cases world wide. I gave an example with 100 to be fair to those who have not reported some issues. If your design can survive tens of thousands, if not hundreds of thousands, of people and only have 50-100 idiots cause one hyper specific issue, you did a great job.

all PC plugs are cheap and very basic design . you "can" make idiot-proof plugs. but no one would pay $20 more for a plug ...
 
Gigantic cards in too small of a case is another huge factor here. Could be as simple as constant stress pushing on the power cable which in turn pushes on the GPU pcb (which probably has a lot of sag too) causing a heat build up in the card that leads to softening the plastic of the connector and leads to the wires/pins moving under the stress. With the heat damage covering up any evidence.

Most cards prior to the 30 series had ample room for the 8-pin connectors to not hit the side panels of cases, so a lot less instances of melted connectors.
I wish they would either make the power connector come straight out of the PCB at a right angle, or make/offer a right angle adapter.
 
Why not just make the top half of the connector one solid metal blade rather than multiple pins, and then do the same for the bottom, then have the card use multiple metal spring contacts that run the length of each blade. This has worked with server power supplies without much trouble, and allows them to maximize the contact surface area for a given size.
2J3hYqX.jpg


Another thing that can be done, are taking design enhancements developed in 1904. The standard outlet specifies a length that is longer than needed based on the contact surface area. The design is to maintain safety even if there is variance in the outlet or plug position. where you can safety pull the rated power, even if the the position is not 100% perfect, as well a taking into account that manufacturing variance exists. The cost to make a plug slightly longer is negligible, and the amperage handling capability ends up being based on the the contact surface within the receptacle.
 
Last edited:
Why not just make the top half of the connector one solid metal blade rather than multiple pins, and then do the same for the bottom, then have the card use multiple metal spring contacts that run the length of each blade. This has worked with server power supplies without much trouble, and allows them to maximize the contact surface area for a given size.

Apple actually did this in the last Intel Mac Pro. They ran power through the motherboard to each GPU. They had their own special GPUs, and for non-Apple products PCIe cables that plugged into the motherboard.


The downside was that the cards had to be a lot bigger and the motherboard was just as wide and had to have a massive power plane built into it.
 
Nothing prevents a person from using all the burners and oven at the same time, so the rating is accurate. US household 240V Ovens are usually connected to a 50A or 60A circuit depending on when the house was built. Perfectly adequate for a roughly 40 amp expected load. Pretty much how my EV charger is set up as well. 40Amp charger on a 50Amp breaker. That higher voltage does allow the use of smaller wires than our typical 120V circuits since the current stays lower. Lower the voltage, higher the amps, worse for wire heating. Which is why a relatively thin AC power cable is thinner than the 8+ wires we use for powering up GPUs and CPUs.
I presume they were talking about the coil used in the induction stove element themselves, not the actual cable used to power the stove. Then again, they didn't specify where these thin wires were coming from. They could've been out of a control board for all I know and they were just trying to be smug.

If the cable powering the stove itself is 5mm x 1.5mm, I'd run away from it. 10kW even out of a 240V socket is nearly 44A. I've handled a 50A cable before, and those things you can use as a blunt force trauma weapon.
 
I presume they were talking about the coil used in the induction stove element themselves, not the actual cable used to power the stove. Then again, they didn't specify where these thin wires were coming from. They could've been out of a control board for all I know and they were just trying to be smug.

If the cable powering the stove itself is 5mm x 1.5mm, I'd run away from it. 10kW even out of a 240V socket is nearly 44A. I've handled a 50A cable before, and those things you can use as a blunt force trauma weapon.
internal wires 5x1.5mm which connects to external cable 5x 2.5mm (H05VV-F 5G2,5)
its 3x230V/16A +1PE +1xN
 
Apple actually did this in the last Intel Mac Pro. They ran power through the motherboard to each GPU. They had their own special GPUs, and for non-Apple products PCIe cables that plugged into the motherboard.


The downside was that the cards had to be a lot bigger and the motherboard was just as wide and had to have a massive power plane built into it.

with Gen 5 PCIe bandwidth , it can be done with shorter motherboards , make the GPU card 8 lanes as standard and the rest length for power ...
 
LOL so would properly seating the connector. The 12VHPWR connector can easily carry 3x or more power and still not fail.

Again search .. PCIe melting .. look at the images. The excess heat is caused from improper seating and resulting from arcing.

The 'RTX 50' series might add a second 12VHPWR connector. Otherwise imagine how insane 8 x 8-pin connectors would be nevermind the risk of failure. The connector is nothing that Nvidia designed.
You very obviously know absolutely nothing about electronics or even electricity.