I actually did exactly this the first day of testing my 4090. Thinking i plugged it in, but it was still like 1mm from being completely plugged in, and it was REALLY hard to get it all the way in. Imo it's just a sh1tty connector, and i still really don't see the point of it... it's just a worse, more unsafe solution than 8 pin pcie cables.

Astyanax:

plastic can't be perfectly molded every time, thats why the design has its specified tolerances, get stronger and push it in right instead of passing off responsibility. hey, maybe blame the AIB's for not using the vendor nvidia told them to use, since FE's, Palit and PNY cards have no issue accepting any adapter or native cable instead of the actually fine connector in the adapter itself. Good man, because even the ATX24, EPS and PCIE 8 pins can melt this way if the connection is bad, and we've seen plenty of those cases.

Dont you ever get tired of shilling for nvidia... we sure do...

Okay, let´s go, let us increase speculation even more: I bet NO, it is NOT the 12VHPWR connector. It is because of very, very, short + very, very, high random power peaks (>1kW/ milisecond), caused by bad quality components on pcb or circuit design, or I don´t know. It´s a problem, you can´t measure with ordinary instruments and it is almost impossible to reproduce. Most good quality PSUs no matter which performance (500W- 1.5kW, low ripple) can withstand these milisecond "power shocks", if it should get critical, actually they shut down (fuse). During a certain time interval (there can be seconds or even minutes in between), If power peaks coming too often and too short in succession the system will break. System will break at most weakest point, in this case the connector, no matter at PSU side or at GPU side. Any other bets?

Astyanax:

plastic can't be perfectly molded every time, thats why the design has its specified tolerances, get stronger and push it in right instead of passing off responsibility.

Connectig one molex to another when you are holding both in your hands may be fixed gettin stronger. You shouldn't "need a hammer" to connect a GPU or any other device. If excessive force is applied it may break other weak points on the GPU (socket, soldering points, etc.)

Goiur:

Connectig one molex to another when you are holding both in your hands may be fixed gettin stronger. You shouldn't "need a hammer" to connect a GPU or any other device. If excessive force is applied it may break other weak points on the GPU (socket, soldering points, etc.)

Yeah, i was honestly quite concerned if i was going to break something with the force i needed to use to get it all the way in.

Far from convinced by anything Corsair say - company has gone down the tubes recently, and I won't be shopping with them again.

Astyanax:

... Good man, because even the ATX24, EPS and PCIE 8 pins can melt this way if the connection is bad, and we've seen plenty of those cases. https://ip.bitcointalk.org/?u=https%3A%2F%2Fi.imgur.com%2FmkTKa.jpg&t=642&c=H_IiWUYnrbtS4w

Look where that connector failed. The place were the design is F awful. If you design something that gives place to user failure, failure will happen. Thats bad design.

Well I too agree that the design is questionable for me as a user, putting a lot more current over less contact space / cross sectional area. At least common sense gives one the impression of this, be it true or not. (And I don't think that the old cables were over dimensioned for safety reasons). Then again, if you build an idiot proof world, you will get more idiots than if you don't. And that's also a fact. All of this just reminds me of my first days of building PCs, always anxious about motherboard connectors being plugged in properly because they're a b**** to do, and even more so, cursing when trying to get them out (because they're usually tight and bothersome to undo).

Can´t see it realy on pictures, but somebody knows, if it is always the same pin/pins or at the same place/ point?

GO figure 🙄 Be an PC owner is to know Your thingy (very well) 😛

So I suppose you dont´t know it either, but tx for great help.

brogadget:

Can´t see it realy on pictures, but somebody knows, if it is always the same pin/pins or at the same place/ point?

its always the active row, not always the same pin(s)

That photo on JohnnyGuru's site is weird. Looks like the main part of connector is fully in, but sense wire pins are out by 1mm. Anyway the part that is melting is the main 12 pin part, not the 4 pin sense part.

Kind of sad... Corsair knows JohnnyGuru has a widely trusted opinion, so, they use him to say "look, everything's fine, really".

Astyanax:

stop blaming the connector, which has been tested for years under the required loads and then some, with flying colours, its obviously a protection method for fragile egos, but doesn't correct the issue going forward.

Yes, under the required loads, but people are overclocking their GPUs to a point where it's exceeding that load. Whether the connector is "only" 99% slotted or perhaps bent a little too far, it is a design flaw if it can heat up this easily. People are stupid, not every computer operates in a controlled lab, and/or mistakes happen: when you're building a connector that can supply enough juice to power some people's entire entertainment system (speakers, display, sound, PC, etc), there really ought to be a lot more room for error than what was given. This connector was a failure. As I've stated in another thread, they should have gone with plates rather than pins. You don't have to be so precise with plates. The added surface area helps dissipate heat. You can more cheaply manufacture the GPU by using thinner gauge wires but more of them. This is proven to be effective in servers.

Astyanax:

Good man, because even the ATX24, EPS and PCIE 8 pins can melt this way if the connection is bad, and we've seen plenty of those cases.

Why do you think Intel is trying to phase out the 24 pin ATX connector? It was just a mild evolution to the 20 pin, where the most power-hungry AIB would maybe be 50W. Most AIBs back then ran cool enough to not even need a heatsink. PCIe then came around and the extra bandwidth meant more processing capabilities, which meant more power draw. So, the extra 4 pins were really just there to help compensate. If you see the connector melted, like in the picture you provided, that shows it was pushed way beyond spec. So long as the connector wasn't one small yank from being pulled out, you're doing things wrong if you're drawing that much power from it. I have a BOINC rig where I've got a pair of <75W GPUs connected to PCIe x1 slots. Aside from the fact I don't want to modify the motherboard to fit a x16 card, I also don't at all trust that the 24 pin connector is going to provide enough juice to power those GPUs, in addition to everything else connected. So, I got one of these to help prevent a fire hazard: https://www.moddiy.com/product_images/f/234/PCI-Express_PCI-E_1X_to_16X_Riser_Card_Flexible_Ribbon_Extender_Cable_w_Molex___Solid_Capacitor__89556_zoom.png I can run these GPUs for days at a time and all of the 12v connectors are cool to the touch.

Astyanax:

its always the active row, not always the same pin(s)

yep, thanks a lot, looks like happens on soldered act. side only. I have not seen any adapter which is crimped only. To be honest, I have never seen a soldered power connector before. But in this particular case, that doesn't mean crimping is the better solution, realy very difficult.

schmidtbag:

Yes, under the required loads, but people are overclocking their GPUs to a point where it's exceeding that load. Whether the connector is "only" 99% slotted or perhaps bent a little too far, it is a design flaw if it can heat up this easily. People are stupid, not every computer operates in a controlled lab, and/or mistakes happen: when you're building a connector that can supply enough juice to power some people's entire entertainment system (speakers, display, sound, PC, etc), there really ought to be a lot more room for error than what was given.

There isn't any exceeding of loads, the connector is overbuilt, the adapter is overbuilt and the issue is happening on stock cards.

schmidtbag:

Why do you think Intel is trying to phase out the 24 pin ATX connector?

They want to make their board partners more money by charging more for mainboards that doo all the regulation at the mainboard level instead of the psu level.

so the guy who works at Corshair making PSU's says its not the connector on the PSU's its user error. this stinks to me of them getting out ahead of the warranty claims . your faulty PSU just trashed my $1500 Graphics card type claim

Astyanax:

There isn't any exceeding of loads, the connector is overbuilt, the adapter is overbuilt and the issue is happening on stock cards.

If it was overbuilt then there is no possible way this would happen. Perhaps the connector would get hot, but it should never burn so long as the connector actually clicked into place. If it were overbuilt, it ought to handle a bare minimum of 1KW. Think of it like airplane tires - they're sometimes filled up to 80PSI, they often experience 200PSI when landing, but they're engineered to have a max of 800PSI. Despite this, they're known to go into the thousands before rupturing. THAT is how you overbuild something. If you seriously think a high-wattage (for a home PC) connector whose maximum rating is 662W is being "overbuilt", then I sure as hell hope you're not an engineer, seeing as people have either approached or already exceeded that. To be overbuilt means there isn't any realistic situation where it could be exceeded.

They want to make their board partners more money by charging more for mainboards that doo all the regulation at the mainboard level instead of the psu level.

That's absurd. Why would Intel give a crap about any of that? The only motherboards they manufacture these days (if they even still do) would be for servers and NUCs, neither of which are affected by this. Intel also doesn't manufacture PSUs. So, regardless of which component is involved, Intel isn't making a dime off this change. They're doing it because it's just a lot simpler and easier. ATX is archaic - we don't need -12v or -5v anymore. I'm not aware of any 3.3v device that is especially power hungry, where the cheapest of components could get the job done just fine. Motherboards will get more expensive but really only based on the features it has. For example, you won't need to supply 30A of 5v if you've got an ITX motherboard and USB 3.1 ports. A 12v-only PSU can be made much smaller and, if built efficiently enough, can be passively cooled.

schmidtbag:

If it was overbuilt then there is no possible way this would happen. Perhaps the connector would get hot, but it should never burn so long as the connector actually clicked into place. If it were overbuilt, it ought to handle a bare minimum of 1KW.

Nonsense, plug a properly built 110 ac plug in incompletely and watch it melt with inconsequential load given enough time. the adapter is certified for 9.5a but material capable for 21a, reaching 1500w without melting or failing. H = I2 x R x T more importantly, in the case of poor connection I2R = ασ2πrl (T4−T40)

schmidtbag:

That's absurd. Why would Intel give a crap about any of that? The only motherboards they manufacture these days (if they even still do) would be for servers and NUCs, neither of which are affected by this. Intel also doesn't manufacture PSUs

🙄 seriously.

Astyanax:

Nonsense, plug a properly built 110 ac plug in incompletely and watch it melt with inconsequential load given enough time.

I'm really not understanding you here. Are you using that as an example of what would never happen? Because I don't see how that's comparable.

the adapter is certified for 9.5a but material capable for 21a, reaching 1500w without melting or failing.

It doesn't matter what it's certified for if manufacturers aren't making it to that spec, which is obviously the case considering people are seeing burning connectors at less than half the limit.

🙄 seriously.

Nice rebuttal there. Clearly, you are the mastermind behind computer component engineering and marketing.