News RTX 4090 Woes Get Worse: Native 16-Pin Reportedly Melts as Well

[Admin]

One GeForce RTX 4090 owner has reported the first meltdown of a native 16-pin power connector from an ATX 3.0 power supply.

RTX 4090 Woes Get Worse: Native 16-Pin Reportedly Melts as Well : Read more

 

[helper800]

It was inevitable...

 

[peachpuff]

Nvidia should get into the home insurance business...

 

[TechieTwo]

The likely issue in this instance is electrical resistance caused by insufficient pin-to-socket contact surface area for the current being consumed.

 

[blacknemesist]

I'm starting to believe something here is fishy.
Sure the adapter might be terrible but the board itself has a bad connector? That seems like too much of a mistake even for NVidia not to test their own boards.. but then again its NVidia so they might have just cut corners even further to capitalize on profits, it backfired massively if that's the case, imagine replacing each and every 4090 and the 4080 having the same issue if its on the board.
I already have my 4090 coming at MSRP but I might just sell it and go for AMD if the benchmarks make it a GREAT value for 4k(Honestly though it because RT will tank them probably).

 

[eye4bear]

This sure brings up the question "who tested this thing and who signed off on it"? They both deserve a "let's shoot ourselves in the foot" marketing excellence award.

 

[-Fran-]

At this point, I think nVidia is getting closer and closer to a massive recall which will cost them millions. Even going back to the 8pins is going to cost them millions right now.

I'm guessing nVidia is just hoping there is no "it lit up and burned my PC" reports, so they aren't forced to do so.

Man... I also want to do so many hot takes, but they make themselves. Smokin' hot takes even!

AMD could not have had better luck with their timing... Mostly luck, for sure, and a little bit of foresight.

Regards.

 

[InvalidError]

The connector size isn't a problem. Enforcing current balance between pins to prevent massive load imbalance if some pins make significantly worse contact than the others is. If you don't want to worry about current balance, then use something like an XT60 - still pretty small with only two pins and #10 wires for up to 55A if you don't mind the ~5W loss per wire.

I'm guessing nVidia is just hoping there is no "it lit up and burned my PC" reports, so they aren't forced to do so.

As long as Nvidia's stuff meets the 94V-0 flammability spec, starting a fire should be practically impossible. The connectors may melt, smoulder and stink up the room but they won't catch fire.

If Nvidia, AIBs, PCI-SIG, Intel (ATX 3.0) and PSU manufacturers cannot find the root cause of these melting connectors and issue definitive fixes for whatever issues are identified, then the HPWR standard will need to get scrapped altogether.

 

[bigdragon]

So how much longer until the class action lawsuits and recalls start? That's got to be coming up soon. Alternatively, we could see a driver or firmware update that reduces 4090 performance and power consumption. I doubt Nvidia will sit idly by as more connectors melt and more negative PR spreads.

 

[jasonf2]

The connector size isn't a problem. Enforcing current balance between pins to prevent massive load imbalance if some pins make significantly worse contact than the others is. If you don't want to worry about current balance, then use something like an XT60 - still pretty small with only two pins and #10 wires for up to 55A if you don't mind the ~5W loss per wire.


As long as Nvidia's stuff meets the 94V-0 flammability spec, starting a fire should be practically impossible. The connectors may melt, smoulder and stink up the room but they won't catch fire.

If Nvidia, AIBs, PCI-SIG, Intel (ATX 3.0) and PSU manufacturers cannot find the root cause of these melting connectors and issue definitive fixes for whatever issues are identified, then the HPWR standard will need to get scrapped altogether.

Agreed. For the standard to exist they are going to have to go to a per pin amperage watchdog and a bios error when the ceiling is hit. It won't fix it, but will keep it from melting. If not they will have to move over to a 2 conductor + ground setup with a big enough connector and conductor set to carry the full load. At what point in this whole mess are is extending the ATX spec with a 24vdc rail starting to make sense rather than trying to deal with 12vdc amperage? It doesn't sound like these cards are going to go down in load anytime soon.

 

[Nikuuuuu]

A GeForce RTX 4090 owner (via Reddit (opens in new tab)) has reported the first alleged case of a 12VHPWR power connector meltdown from a native ATX 3.0 power supply.

Just to let you know, the user on the Reddit posts aren't the the ones with the connector issue. It was a repost of someone else who is having the issue on Facebook.
Here is the proof along with the link of the original post.

 

[InvalidError]

At what point in this whole mess are is extending the ATX spec with a 24vdc rail starting to make sense rather than trying to deal with 12vdc amperage? It doesn't sound like these cards are going to go down in load anytime soon.

The PCI-SIG has already introduced an optional 48V HPWR spec, may as well go with that. USB 240W is a 48V-only spec too.

 

[Kamen Rider Blade]

According to the updated Reddit post.

The user of the 4090 was using the MSI MEG Ai1300P power supply .
That PSU was using substandard cable gauge.

Spoiler: Burnt Native Plug

MSI was using 18-24 AWG on the 600W 12VHPWR 12+4 PCIe .

Spoiler: Cable Gauge list for MSI PSU MEG Ai1300P

nVIDIA's 12VHPWR Adapters that they contracted out to "Astron" to Design/Manufacture was using 4x 14 AWG copper wire pairs to the 4x PCIe 6/8pin Receptacles.

CableMod Cables were using 6x 16 AWG copper wire pairs in their adapters.

 

[TJ Hooker]

The connector size isn't a problem. Enforcing current balance between pins to prevent massive load imbalance if some pins make significantly worse contact than the others is.

If one or more of the connectors is making poor contact, would current balancing run the risk of causing that connector(s) to get hot? Because the same amount of current would be forced to go through it despite the increased resistance.

 

[TJ Hooker]

According to the updated Reddit post.

The user of the 4090 was using the MSI MEG Ai1300P power supply .
That PSU was using substandard cable gauge.

Spoiler: Burnt Native Plug

MSI was using 18-24 AWG on the 600W 12VHPWR 12+4 PCIe .

Spoiler: Cable Gauge list for MSI PSU MEG Ai1300P

nVIDIA's 12VHPWR Adapters that they contracted out to "Astron" to Design/Manufacture was using 4x 14 AWG copper wire pairs to the 4x PCIe 6/8pin Receptacles.

CableMod Cables were using 6x 16 AWG copper wire pairs in their adapters.

Wow. The PCIe 5.0 12VHPWR spec calls for 16 AWG (excluding the sense pins). MSI was explicitly billing this as the first fully ATX3.0/PCIe 5.0 compliant PSU on the market, and they don't even comply with the cable specs?

Edit: Looks like the Cybenetics spec sheet for that PSU has been updated, now lists the 12VHPWR connector as being 16AWG. Maybe the original listing of 18AWG was a typo?

 

[Kamen Rider Blade]

One thing that is good about 12VHPWR standard, it's exposing shoddy Electrical work in PSU / Wire / Adapter manufacturers.

Wow. The PCIe 5.0 12VHPWR spec explicitly calls for 16 AWG (excluding the sense pins). MSI was explicitly billing this as the first fully PCIe 5.0 compliant PSU on the market, and they don't even comply with the cable specs?

MSI was trying to pull a "Fast One", this wouldn't be the first time in MSI's history that they tried to do something like that.
Look at the wiring report, it's clear as day.

The PCI-SIG has already introduced an optional 48V HPWR spec, may as well go with that. USB 240W is a 48V-only spec too.

Given the turbulent, flammable rollout of 12VHPWR so far, how do you think the wider consumer end of the PC community will receive 48V HPWR?

I don't think anybody wants to adopt any new specs until the 12VHPWR debacle is one & done for over a decade or two at least w/o issues.

 

[InvalidError]

If one or more of the connectors is making poor contact, would current balancing run the risk of causing that connector(s) to get hot? Because the same amount of current would be forced to go through it despite the increased resistance.

If resistance on one pin increases, current through that wire will decrease, which would force the current-balancing circuitry to reduce current across all other wires as well. From there, the GPU can either reduce its power limit to match what the compromised connector is still able to provide before the 12V rail shows excessive droop or hit the VRM's under-voltage lock-out threshold and stop altogether.

Another option would be to individually monitor current and voltage on each pin.

Either way, that would be a fair amount of extra circuitry specifically for connector protection. It would be much simpler to go XT60-like with 2x#10 wires.

Given the turbulent, flammable rollout of 12VHPWR so far, how do you think the wider consumer end of the PC community will receive 48V HPWR?

Most consumers don't give a damn since they buy their computers off the shelf. OEMs could go 48VO and most wouldn't notice since they'll never open the case. I would be fine with the whole industry making the jump to 48V. As I have written before, adapters for pre-xxVO will be a thing for people who want to carry over some legacy stuff like SSDs and HDDs.

 

[helper800]

Nvidia should get into the home insurance business...

They should bundle it with every 4090.

 

[mlambert890]

I'll worry about it when it's more than a handful of cases on Reddit, out of thousands sold, with confirmation bias amplifying it to the max, and the media jumping on the hype bandwagon.

Not saying there isn't some issue, but the level of "doom" hysteria is, as usual, ridiculous given the amount of data points. I know feeding this kind of frenzy is lucrative "journalism", but it's always sad to see.

 

[jasonf2]

The PCI-SIG has already introduced an optional 48V HPWR spec, may as well go with that. USB 240W is a 48V-only spec too.

That keeps it just in low voltage range and should give plenty of headroom to power up to 700 watts on two short 14 ga conductors (15 amps). Not that there is probably going to be a huge adoption of that charging standard on computer ports, but it at least gives some additional purpose. If they don't change something up voltage loss problems and over amping connectors are going to become more problematic. The biggest uptake issue with 48v (or 24 for that fact) would be defato requirement for a whole new power supply with the new power rail and dedicated connector(s).

 

[jasonf2]

I'll worry about it when it's more than a handful of cases on Reddit, out of thousands sold, with confirmation bias amplifying it to the max, and the media jumping on the hype bandwagon.

Not saying there isn't some issue, but the level of "doom" hysteria is, as usual, ridiculous given the amount of data points. I know feeding this kind of frenzy is lucrative "journalism", but it's always sad to see.

Product liability cases like this are the stuff that class action attorneys dreams are made of. The 4090 card isn't the real issue here. The new header is destined to be used on millions of cards and even if this is only happening on a fraction of a percent of all cards involved the product liability could run in the billions if it starts fires. With all of the press this has already gotten it is very much in the interest of all parties that would use this connector to fuse the pins on the connection either mechanically or electronically. Going beyond that even if only a fraction of a percent are melting down, that is enough to say I wouldn't put one of these in a new build until the issue is fixed with at least a firmware update. Murphy's law applies.

 

[cryoburner]

It's almost as if operating a small space heater inside your PC isn't a good idea. : 3

As long as Nvidia's stuff meets the 94V-0 flammability spec, starting a fire should be practically impossible. The connectors may melt, smoulder and stink up the room but they won't catch fire.

I also suspect most PC cases would prevent it from spreading further, even if components inside the case managed to ignite. Something like an open-air case might be risky though.

I'll worry about it when it's more than a handful of cases on Reddit, out of thousands sold, with confirmation bias amplifying it to the max, and the media jumping on the hype bandwagon.

Not saying there isn't some issue, but the level of "doom" hysteria is, as usual, ridiculous given the amount of data points. I know feeding this kind of frenzy is lucrative "journalism", but it's always sad to see.

The card's only been out a couple weeks though, and there's no telling how many connectors might be smoldering away inside people's computers who just haven't noticed it yet. I agree that it will probably only affect a relatively small portion of these high-wattage cards, but that might be enough for recalls of at least some of this hardware.

 

[drivinfast247]

The likely issue in this instance is electrical resistance caused by insufficient pin-to-socket contact surface area for the current being consumed.

No s***!

 

[jasonf2]

One thing that is good about 12VHPWR standard, it's exposing shoddy Electrical work in PSU / Wire / Adapter manufacturers.


MSI was trying to pull a "Fast One", this wouldn't be the first time in MSI's history that they tried to do something like that.
Look at the wiring report, it's clear as day.


Given the turbulent, flammable rollout of 12VHPWR so far, how do you think the wider consumer end of the PC community will receive 48V HPWR?

I don't think anybody wants to adopt any new specs until the 12VHPWR debacle is one & done for over a decade or two at least w/o issues.

I think there is somewhat of a misconception here. Everything up to 50 volts is considered low voltage. Almost all wire is rated above 150 volts (with most in the 300-600 range) and the arcing potential of 48v vs 12 volts is only negligibly more and easily contained in a Molex plug. In many ways a 48V HPWR standard makes a lot more sense for this intended purpose because it gives 4 times the headroom. For every watt used 12 volt requires 4 times as many amps. Amperage is what makes heat, not voltage. So even though 48 volts sounds like it would be more dangerous, in application, it would run significantly cooler with less voltage loss by percentage with no real difference in safety. The biggest barrier here is that power supply cost would go up by adding another power rail. 12 volts downfall here is the need for very low impedance (I suppose it is technically only resistance with it being DC) paths to support the wattage. There is a reason that they use 120v in the US and 220 in Europe for basic household power delivery. You need the voltage to safely deliver the wattage on an economical conductor. I am not a big fan of 12 volt for higher wattage power delivery and there is a reason that 24VDC is as prevalent in the industrial controls side as it is. If you need to go beyond about 15-20 amps voltage loss starts to become an issue and your conductor needs to get big pretty quickly. A 5% voltage sag on 12 volts is only .6 volts and that doesn't give you much to work with at higher wattage.

 

[InvalidError]

The new header is destined to be used on millions of cards and even if this is only happening on a fraction of a percent of all cards involved the product liability could run in the billions if it starts fires.

Components in electronics turn into a charred mess every day, yet you don't hear about fire marshals calling for blanket action to do more about components turning into fireballs. Why? Because such action has already been taken in the past, which is why nearly all officially distributed consumer electronics is made from materials that follow the 94V-0 or similar flammability standard. Short of running your PC with an open side panel side-down on a flammable carpet, faulty connectors won't start any house fires.