The Nvidia RTX 4090’s cable-melting woes are still ongoing, with mounting reports of the 16-pin 12VHPWR power adapter overheating and becoming damaged - in some cases damaging the $1,599 GPU.
A megathread on Reddit is collating documented cases of RTX 4090 adapters that have met with a terrible fate, with the count now rising to 23 confirmed incidents (and six unconfirmed). With the RTX 4080 arriving very soon and bringing the same 16-pin connector with it, some salvation may have appeared in the shape of Corsair’s power supply expert JonnyGuru.
Jonny is well-known in the PSU space as one of the most knowledgeable experts in the field. He’s the latest in a long line of professionals and influencers to try to replicate the RTX 4090’s issues, and has studied the reported cases to come up with a possible solution.
There have been many theories surrounding what causes the 12VHPWR adapter to fail, and Jonny’s belief isn’t a new idea: he’s positing that the adapter’s poor design makes it very difficult to fully insert into the GPU’s power socket.
In multiple recorded cases, Jonny noted that the adapter was improperly inserted; enough to power the card, but potentially not forming a safe connection. The massive size of the RTX 4090 forcing some users to bend their cables may have exacerbated the issue.
While there’s no guarantee that JonnyGuru is correct, his testing process was rigorous, cross-referencing other users’ photographs and using a thermistor to monitor the internal temperature of the adapter.
Analysis: Nvidia is at fault here, but so is the PCI-SIG
As a quick recap for the unaware, the Nvidia RTX 4090 has been having problems with the 12VHPWR adapter, which is a 16-pin PCIe power connector that needs to be used with the new GPU in order to function properly with older ATX 2.0 power supplies.
The PCIe Special Interest Group (PCI-SIG), a consortium of tech manufacturers of which Nvidia is a part, issued a memo earlier this year noting that the 12VHPWR adapter could have safety issues involving ‘thermal variance’, but Nvidia put the card into production anyway.
Now, people’s crazy-expensive new graphics cards are being burned by the overheating adapters, and the PCI-SIG has released confirmation that a revision to the 12VHPWR design may be coming in December, as shown in the tweet below.
ECR = engineering change requesthttps://t.co/GxmpjzA4pEVery interesting, what will be changed at this time🧐🧐🧐 pic.twitter.com/iFOioScvklNovember 8, 2022
What the PCI-SIG's governing body hasn't done is admit any kind of responsibility here - and it looks like Nvidia isn't willing to either. In our view, the blame sits fairly evenly between the two parties; Nvidia manufactured the cards, but the PCI-SIG didn't properly act to prevent the issues we are starting to see.
As an aside, if you’re trying to get your hands on an RTX 4090, you can potentially bypass this whole cable-melting affair by pairing the GPU with a newer ATX 3.0 power supply rather than one of the ATX 2.0 models that are still more common among PC-builders.
JonnyGuru has one last bit of advice (beyond ‘make sure the connector is fully inserted’) - he believes that a small amount of dielectric grease, which can be purchased cheaply on Amazon or from tech hardware retailers, applied to the connector pins can supposedly improve connectivity.
Nvidia has yet to make an official statement regarding the thermal issues, though we do know that Team Green is looking into the problem and has contacted several affected individuals to confirm that an investigation is ongoing.
