Ongoing Melting Issues with NVIDIA RTX 4090 Graphics Cards: Repair Shops Manage 200 Units Monthly

Published by

Click here to post a comment for Ongoing Melting Issues with NVIDIA RTX 4090 Graphics Cards: Repair Shops Manage 200 Units Monthly on our message forum
https://forums.guru3d.com/data/avatars/m/258/258465.jpg
I'm so glad that I've bought the 4080 when I read this 😀
https://forums.guru3d.com/data/avatars/m/300/300104.jpg
Hopefully the 5090 resolves this issue, however I feel a lot of it could be user error.
https://forums.guru3d.com/data/avatars/m/217/217682.jpg
It's 100% user error or crappy cablemod adapters. Use ONLY adapters from PSU manufacturers like Corsair or Seasonic. Personally I use Corsair's 12VHPWR cable(2*8pin to 16pin) and everything is fine.
https://forums.guru3d.com/data/avatars/m/258/258664.jpg
Yeah... 5090 won't have this issue since I bet they just slap two connectors on it, being safe again with less power running over individual wires. That said, I'm still waiting to see if I need the firefighters at my place anytime soon. First 3 months seemed to be okay so far.
https://forums.guru3d.com/data/avatars/m/227/227994.jpg
fantaskarsef:

Yeah... 5090 won't have this issue since I bet they just slap two connectors on it, being safe again with less power running over individual wires. That said, I'm still waiting to see if I need the firefighters at my place anytime soon. First 3 months seemed to be okay so far.
Yep, two of them will solve the problem. From what i remember (from Der8auer) one connector is on the limit with a maximum of 615 watt or something.
https://forums.guru3d.com/data/avatars/m/248/248994.jpg
Conan The Barbarian:

Hopefully the 5090 resolves this issue, however I feel a lot of it could be user error.
It's a user error propagated by a poor design of the connector. The connector was apparently perfect in a laboratory but nobody tested it in the wilds before it appeared in commercial products. It wouldn't surprise me if an AI was used to design it. The same AI that draws people with six fingers per hand regularly.
https://forums.guru3d.com/data/avatars/m/220/220214.jpg
Does this also affect any 4080 cards with this bad connector or it is just the higher powered 4090 it affects?
https://forums.guru3d.com/data/avatars/m/227/227994.jpg
geogan:

Does this also affect any 4080 cards with this bad connector or it is just the higher powered 4090 it affects?
Seems to be 4090's mainly. There have been melted 4080 connectors, but only when using that CableMod Adapter. A regular cable should be fine.
https://forums.guru3d.com/data/avatars/m/246/246564.jpg
4080s don't have the power draw to melt the cable. As long as you're using the supplied adapter or your native PSU one, you'd have to go out of your way to melt it.
data/avatar/default/avatar25.webp
Kaarme:

It's a user error propagated by a poor design of the connector. The connector was apparently perfect in a laboratory but nobody tested it in the wilds before it appeared in commercial products. It wouldn't surprise me if an AI was used to design it. The same AI that draws people with six fingers per hand regularly.
And you can blame Intel for that, as they are the ones who proposed the original 12VHPWR connector to the PCI Express forum. You can also blame the PCI Express forum for doing little to no stress testing before officially adopting this connector as the new power supply standard connector for GPUs, and finally you can blame NVIDIA for going with this standard on the 40xx series cards without doing their own extensive testing (I assume that they used open bench setups for their own testing, which probably avoided the stress on the connector when plugged into the GPU.
https://forums.guru3d.com/data/avatars/m/226/226864.jpg
Memorian:

It's 100% user error or crappy cablemod adapters. Use ONLY adapters from PSU manufacturers like Corsair or Seasonic. Personally I use Corsair's 12VHPWR cable(2*8pin to 16pin) and everything is fine.
The cable that came with one of our ASUS PSUs, for example, gradually slipped out of the card by itself, the cable that came with the other PSU of the same make was fine, as were cables and 180° adapters (we used them until the recall) from Cablemod. An acquaintance who doesn't know much about computers had me check her son's rather new PC with a 4080 Super after it crashed and she smelled burning had a PSU with native 12VHPWR cable (12VHPWR connector on both sides) which melted on the PSU side. The shop that had built the PC exchanged the PSU and her computer has been running fine ever since. So you aren't "safe" even with the cable from your PSU manufacturer. Also, I don't think it is user fault. There may be some instances where the cable isn't clicked in fully or bent too hard too close to the connector, but I know someone with a 4090 that used a BeQuiet native PSU cable (he even got a case where he didn't have to bend the cable very much) and checked every day before he started up his PC that the cable was still fully seated (he was really worried about all the reports of melting) and clicked in and the connector melted anyway. He only had the card for about a month or two. I myself, on the other hand, had my card since shortly after launch and it's thankfully still okay, even though I had used a faulty cable that came with my PSU (the one that slipped out by itself) for a while. I also had very low 12VHPWR voltage (even going down to 11V under full load) with various cables I used earlier (including the one that came with the PSU and another PSU, an early cablemod adapter and the adapter dongle that came with the card) until I switched to the Cablemod 180° adapter and straight cable which fixed my voltages and after it was recalled with the Cablemod 90° cable it's fine as well. I hope it stays that way, but who knows. Personally, I don't trust this thing and am doing a visual check with a flashlight to see if it's still in there at least once a day (good thing I have a glass window on the case), physically checking it at least once a week and am also monitoring voltages with HWInfo with an alert in case they drop off too much. It is a bit alarming that I have personally seen the melting happen with the computers of two people I know, especially since not many people I know have cards with this connector yet. I've never known anyone that had other connectors melt before, but it could be coincidence. I think the whole standard is sub-standard with manufacturing with tight tolerances probably being unreliable as well. Even if you don't commit "user error" you still roll the dice on whether you will get affected or not.
data/avatar/default/avatar34.webp
how explain user fault if the same don´t happens with 4080, 4070ti and 4070 cards? these cards sells much more than 4090 and nobody connects wrong him? even cablemod cable connects fully and give a click sound and continues melting
data/avatar/default/avatar07.webp
Kaarme:

It's a user error propagated by a poor design of the connector. The connector was apparently perfect in a laboratory but nobody tested it in the wilds before it appeared in commercial products. It wouldn't surprise me if an AI was used to design it. The same AI that draws people with six fingers per hand regularly.
exactly, mainly when delivers more than 400w, the single 12 pin connector cannot support this
https://forums.guru3d.com/data/avatars/m/126/126739.jpg
Man I would sure be mad, if this happened to my $1800 video card. I'm sure Nvidia wont make the same mistake next gen.
https://forums.guru3d.com/data/avatars/m/239/239175.jpg
Memorian:

It's 100% user error or crappy cablemod adapters. Use ONLY adapters from PSU manufacturers like Corsair or Seasonic. Personally I use Corsair's 12VHPWR cable(2*8pin to 16pin) and everything is fine.
A cousin of mine uses the included 4x8pin to 12VHPWR adapter cable. Which sounds like an accident waiting to happen, IMO.
https://forums.guru3d.com/data/avatars/m/198/198862.jpg
200 users every month making a same mistake? Yeah, user error 😀
data/avatar/default/avatar12.webp
squalles:

how explain user fault if the same don´t happens with 4080, 4070ti and 4070 cards? these cards sells much more than 4090 and nobody connects wrong him? even cablemod cable connects fully and give a click sound and continues melting
I have a theory that rather than power, it's probably because the 4090 cards are so big and chonky that people trying to squeeze them into cases of certain sizes and dimensions tend to bend and yank the power connector from the socket after being fully plugged in thus causing melted connectors. On the other hand, it's also been proven that cablemod 90 degree connectors were problematic causing even more melting. The 3090 FE also had the earlier version 12-pin power connector @450W and we didn't hear too much of a fuss about melting connectors on that one. My 450W 4090 is already 1.5 years old and heavily gamed on and the 12vhpwr connector hasn't melted. Made sure it's plugged in fully and the cable isn't bent too much because I have only installed it in a rather large case with a lot of space for the card, with enough relief for the PSU OEM 12vhpr to 2x8-pin plug and its 750mm cable to not be bent when closing the side panel. Still I agree that the design can still be improved to make it more fool proof but I'm guessing the ATX3.1 PCIE 5.0 12V-2x6 revision of the connectors will probably be featured in the next gen 5090 so I'm hoping those cards won't be so damn huge since the size and weight of the 4090s seem to be problem-prone in more ways than melting, like PCB cracking for instance.
https://forums.guru3d.com/data/avatars/m/275/275921.jpg
mattm4:

Man I would sure be mad, if this happened to my $1800 video card. I'm sure Nvidia wont make the same mistake next gen.
You have nothing to worry about.
https://forums.guru3d.com/data/avatars/m/275/275921.jpg
Undying:

200 users every month making a same mistake? Yeah, user error 😀
1st world problems.
https://forums.guru3d.com/data/avatars/m/108/108389.jpg
So the worst that happen is a melted connector, which can be then replaced. Still better than having to deal with the 3x8pin PCIe cables montrosity