It has been roughly three weeks since the first reports of melting 12VHPWR connectors and NVIDIA says that it is still investigating the matter. A “16 Pins Adapter Megathread” has been created on Reddit showing that the first confirmed case there was reported on October 24. As of November 11, there are now 24 confirmed cases and another 6 unconfirmed. An interesting detail to note is that there are no Founders Edition graphics cards on the lists but major manufacturers such as ASUS, GIGABYTE, MSI, and Zotac all have multiple different models on them. GALAX and Inno3D even managed to make an appearance on the lists. As not all RTX 4090 owners are guaranteed to report on Reddit, nor could be expected to be regularly inspecting their cables and connectors, there could be more incidents of them melting that are not yet known.
There are many theories as to the root cause of these reports but none so far have been consistent enough to fully explain them. On October 27 NVIDIA asked its AIC partners to collect RTX 4090 Cards affected by melted 16-Pin power connectors for its investigation. Kitguru has now reported that an NVIDIA PR person has responded to them with a statement saying that the investigation continues.
“We continue to investigate the reports, however, we don’t have further details to share yet. NVIDIA and our partners are committed to supporting our customers and ensuring an expedited RMA process for them.”
Is user error to blame?
One of the many theories regarding damaged connectors and cables points to user error. Corsair PSU Expert Jon Gerow (Johnguru) issued a report of their own saying the issue is likely user error in not fully inserting the cables into their respective connectors and provided the following images for reference.
MSI has posted the following instructional images on its MSI Gaming Facebook page.
However, some users have stated they did properly install their cables while others have said the connector could not be fully inserted. If accurate, this could indicate potential product defects, but that remains to be seen, and will hopefully be clarified upon the results of the investigation.
Discussion (11 replies)
Join Discussion →One odd thing is that some people are stating that the MSI instructions are incorrect because they say they plugged the wrong end of the connector into the power supply. I'm not sure if there is an incorrect way as I looked at my FSP instructions and nothing is noted.
Yeah, and people on Reddit were joking about the MSI instructions in that they only highlighted the PSU connector and nothing about the graphics cards which is what the majority of the reports have been about.
Maybe that was what they were talking about in the thread I was reading as well. Just up to early and misinterpreted what they meant.
Yeah, about that, that is no longer true.
I guess the so-called "experts" are hiding now! Where do they find these
clownsanalysts?! ?Direct link that doesn't redirect through Google:
[URL unfurl="true"]https://wccftech.com/fully-connected-16-pin-connector-on-the-nvidia-geforce-rtx-4090-ends-up-melting-too/[/URL]
Edit: the site above mentions an engineering change request (ECR) of the 12VHPWR cable by PCI-SIG, and links to the review zone. I'm uncertain of the significance, however.
—https://pcisig.com/specifications/review-zone
No worries. Most sites are just regurgitating stuff with this but I am trying to connect the dots hence why I include extra things and then:
I thought it was odd that no FEs were listed. Sure, they could've had a better connector on the card but if that isn't the issue then it won't matter, and again why I leave my own hooks. I do come from a family of hunters and fishermen.. ;)
It's all good. I posted about this in the AIC thread and went a little deeper for this particular post. The change is for the shroud around the "sense" data pins and is said to not affect whatever is happening with his.
I will say for all concerned, the bit about user error may have some weight. As I have 2 3090 Ti's, and many benchmarks show them using over 450W with a 12-pin to 3x 8-pin adapter. I've been watching them closely. For the hybrid, I don't have to OC it as it already clocks high. For the air-cooled card, I've been gaming, in particular CP 2077, and using Precision X1 for a maximum boost along with the 107% power limit boost, and a mild core/mem OC to force it a bit further. I'm not going to waste time with TPU-Z as Igor also had an article about TBP vs actual power at the wall and so on but I can tell you it's been fine. However, since this is in an open chassis I can walk up and touch it, at the GPU connector, and tell you it's warm to the touch. It has the better part of 70 mm straight to the card and at the PSU. Not that it matters though and not something anyone should be doing (touching while live or with a PSU that hasn't been discharged).
Since this type of connector will "clip/snap" even without fully being inserted, I've witnessed it, this is something to be concerned about. I'm paranoid enough that I worry about such things and even if I hear that click I'm going to inspect it further until I believe everything is right. When I installed my cards I saw this and simply worked with it until I felt it was right. So far, so good.
You really do have to push down extra hard to connect the adapter into the video cards I'm finding, what was typical force prior on 8-pin connectors is not going to cut it now, you have to put even more forced to really make sure it's connected all the way. I do see this as a big issue, I don't think people are doing that, they are used to using a certain amount of force for decades, but now, you have to use more and make sure it's fully flush, it's more effort, and this is a learning and educational thing.
Doing a post right now that includes this and much more info from Igor. He's really gone all-out in researching it and in contact with both NVIDIA and its partners.
[embedded media]
(Haven't watched yet)
Thanks! I'll watch it tonight when I get home.