News Nvidia’s Melting 16-Pin Connectors on RTX 4090: All We Know

[Admin]

Some Nvidia GeForce RTX 4090 16-pin 12VHPWR adapters are melting. There are still more questions than answers, but here’s what we know so far.

Nvidia’s Melting 16-Pin Connectors on RTX 4090: All We Know : Read more

 

[-Fran-]

May want to update and add this as another reference now:

View: https://www.youtube.com/watch?v=ig2px7ofKhQ


Thanks for keeping track on your own as well.

Regards.

 

[cyrusfox]

Gamer Nexus solid video, design error mingled with user error on a card that is $1600.

I get it though. I don't know how many times I have drained and refilled my custom loop, go to boot the computer and forget to plug back in the power cables to the GPU. Easy to relate to those excited to take this for a spin and don't have it fully-seated + cables stretched due to cable management, card boots but is drawing too much current through too little pins leading to overheating and some smoking plastic.

The stories about poor solder joints never passed the smell test, Looking more at aesthetics than current carrying capability.

 

[cloud7s7]

Early access hardware joins early access gaming. Devolution

 

[Gunkk0]

Lines around the block.
Ends justify the means.
Greed wins.
We all lose.

 

[Darkbreeze]

What I'd REALLY like to know is why this hasn't happened on ANY RTX 3090 ti cards that I am aware of, despite them using the exact same 12vhpwr 16 pin connector and having the exact same 450w power draw? I have a lot of respect for Steve at GN, but it would have been nice for him, or anybody, to address and explain the reasoning behind that.

 

[Kamen Rider Blade]

What I'd REALLY like to know is why this hasn't happened on ANY RTX 3090 ti cards that I am aware of, despite them using the exact same 12vhpwr 16 pin connector and having the exact same 450w power draw? I have a lot of respect for Steve at GN, but it would have been nice for him, or anybody, to address and explain the reasoning behind that.

Very few normies were able to buy RTX 3090 Ti's due to the Crypto Mining craze & GPU drought.

Now, alot of normies are buying the RTX 4090.

 

[Darkbreeze]

Well, I've seen plenty of "normies" (And others) on this forum over the last six months or so with those cards in their build and can't recall any of them having this problem. Which, if it's user error, there definitely should have been.

 

[palladin9479]

Well, I've seen plenty of "normies" (And others) on this forum over the last six months or so with those cards in their build and can't recall any of them having this problem. Which, if it's user error, there definitely should have been.

There are multiple failure modes, user inserting halfway is just one of them. It's likely a combination of multiple things happening at once, bad batch of adapters get made that pins don't quite go far enough, batch gets distributed around. Users opens box, installs card and push's cable in thinking it's all the way in, turns on computer and starts having fun. Later the failure occurs in a very unsafe way.

Stuff like this happens all the time and is resolved with the RMA process, what elevates this is the words "consumer" and "potential fire hazard", the device should of failed before getting hot enough to start that kind of damage or had some warning method to let the user know something is wrong.

 

[Darkbreeze]

Yes, I understand all that. Has nothing to do with why we haven't seen it on the RTX 3090 ti with identical 16 pin connector and identical 450w power draw though.

Based on the GN deep dive it would seem that "bad batch of connectors" probably isn't a thing in this particular situation. More like, somewhat poor quality for ALL connectors, at least in terms of the low quality plating. Garbage in, garbage out.

But even that, I haven't seen any signs of from any 3090 ti owners, or heard of, at all.

 

[Deleted member 431422]

Yes, I understand all that. Has nothing to do with why we haven't seen it on the RTX 3090 ti with identical 16 pin connector and identical 450w power draw though.

Based on the GN deep dive it would seem that "bad batch of connectors" probably isn't a thing in this particular situation. More like, somewhat poor quality for ALL connectors, at least in terms of the low quality plating. Garbage in, garbage out.

But even that, I haven't seen any signs of from any 3090 ti owners, or heard of, at all.

I guess that's what Jensen had in mind when he said "... We're going to survive the hard times". Reduce quality to sustain high revenue.

 

[mo_osk]

Yes, I understand all that. Has nothing to do with why we haven't seen it on the RTX 3090 ti with identical 16 pin connector and identical 450w power draw though.

Based on the GN deep dive it would seem that "bad batch of connectors" probably isn't a thing in this particular situation. More like, somewhat poor quality for ALL connectors, at least in terms of the low quality plating. Garbage in, garbage out.

But even that, I haven't seen any signs of from any 3090 ti owners, or heard of, at all.

We haven't "seen" any 3090ti with similar issues. Doesn't mean it hasn't happened. When it started happening to the 4090 people were under the impression that it only happened to power connector adaptators, not the native connectors. You should be more carefull when making assumption.

The 3090 cards were slow to reach the market. Its simply possible that by the time some people had problem with their power connectors we weren't paying as much attention to it as to when the 4090 was just out. The incident would also have been fewer and more apart, it wouldn't give the impression that there was a specific problem. After all, melting power connector happen all the time and people usually assume it was caused by user error or a faulty power supply. Here its only the fact that it seemed to happen systematically that raised the public's suspicion.

ps : also only the FE 3090 have that connector, I don't have any number but I'm sure there isn't that many of them.

 

[JarredWaltonGPU]

What I'd REALLY like to know is why this hasn't happened on ANY RTX 3090 ti cards that I am aware of, despite them using the exact same 12vhpwr 16 pin connector and having the exact same 450w power draw? I have a lot of respect for Steve at GN, but it would have been nice for him, or anybody, to address and explain the reasoning behind that.

The one sample I received (from Asus) uses a 12-pin adapter, without the four extra sense pins. That's logical, as the sense pins aren't going to do anything useful with an adapter. They're for communication between the PSU and the destination 16-pin card, and with an adapter there's no meaningful receiver of information. So, my thoughts are that the extra sense pins make it just a bit more difficult to insert the 16-pin compared to the 12-pin used on all the RTX 30-series Founders Edition cards and the 3090 Ti (AFAIK).

 

[thently]

View: https://www.youtube.com/watch?v=_QmKYJzJhB4


More info from Nvidia now.

 

[lmcnabney]

There are most likely an order of magnitude number of these incidents, but most users just think their GPU failed and RMA it without carefully inspecting the inside of the power connection.

 

[TJ Hooker]

There are most likely an order of magnitude number of these incidents, but most users just think their GPU failed and RMA it without carefully inspecting the inside of the power connection.

Sure, the user may not realize, but once they RMA it I assume the manufacturer will examine the card and see what happened. So Nvidia should be able to get a pretty good idea of the numbers (with some lag time).

 

[lmcnabney]

Sure, the user may not realize, but once they RMA it I assume the manufacturer will examine the card and see what happened. So Nvidia should be able to get a pretty good idea of the numbers (with some lag time).

Has any tech company ever publicly released questionable RMA numbers on their own products? I'm sure that Nvidia has a much clearer understanding of the situation, but it is marketing and PR that respond to issues like this - not the engineers.

 

[TJ Hooker]

Has any tech company ever publicly released questionable RMA numbers on their own products? I'm sure that Nvidia has a much clearer understanding of the situation, but it is marketing and PR that respond to issues like this - not the engineers.

Probably not, but in that case I imagine they simply wouldn't release failure/RMA rates at all. Knowingly publishing false numbers would be different, and I think might even open them up to a potential lawsuit if they are discovered.

 

[Darkbreeze]

It's basically a moot point, at this point. Already diagnosed. And seemingly, no further "incidences" since identifying the lack of plugging things in properly as the root source of the problem. Not sure, unless further evidence comes up, there is any reason to even talk about this further at this point.