The problem of overheating to the point of melting the 12-pin 12VHPWR connectors with the new Nvidia graphics does not stop, but rather spreads. It turned out that it not only affects 8-pin adapters, but the native cable directly from the source can also melt.
Unfortunately, it seems that soon there will be no peace from the matter of the 16-pin ATX 3.0 12VHPWR power connectors, which in some cases burn or melt in the GeForce RTX 4090 due to overheating of the contacts in the connector. Apparently it’s not just a specific type of adapter that’s problematic for the quality of workmanship, there have already been cases where the 12VHPWR cable connected directly to the source has melted. And there are also new tips on how to defend yourself.
It is still not entirely clear if the 12VHPWR connector is really problematic from a systemic point of view, that is, if it will always be prone to these problems, or if some revisions in the design and in the quality and in the manufacturing precision will improve the reliability to such a level. that overheating will no longer occur in practice. However, cases of overheating are not yet disappearing. Previously it was reported that Nvidia asked card makers to take back the hardware components affected by users, including power adapters (which they probably should have anyway due to the warranty) and then hand them over to Nvidia for review. The existence of the problem is therefore accepted as real, at least to the extent that some investigations are underway.
Moreover: Warning: 12-pin connectors on GeForce RTX 4090 cards can burn. How to prevent problems?
Gainward should have changed the adapters, but ultimately won’t
In addition to this internal Nvidia decision, information about another has appeared. Gainward, one of the card makers, is said to have decided to retire the power adapters from the eight-pin connector to the 16-pin (or 12 + 4-pin) connector that it supplied to its GeForce RTX 4090 cards. autonomous decision of this company, not an instruction from Nvidia.
However, this information has since been refuted, the company has released a statement that it won’t change the cables. According to Nvidia, it inspected the supplied adapters and found no problems.
Other than that, things are getting wired. According to someone with a dead adapter (Gigabyte card), NVIDIA is blaming Gigabyte. But as far as I know, NVIDIA doesn’t just supply the chips. Partners must purchase to complete adapters. It should still be NVIDIAs take.
– Andreas Schilling 🇺🇦 (@aschilling) November 7, 2022
Andreas Schilling of the HardwareLuxx website makes an interesting point: Third-party cards should also have Nvidia-supplied power adapters. Presumably it requires that, in addition to the GPU, the card manufacturer also purchase the adapters included with the cards from it. However, this would mean that even for adapters that may have caused problems with partner manufacturers’ cards, Nvidia could be directly responsible. While it is possible that this information will change somewhat, for now it only comes from one source.
The problem is not caused by the adapters themselves, it can also occur without them with a cable directly from the source
After the first incidents, there was an opinion that only 12 + 4 pin adapters to traditional eight pin connectors cause problems, and with a “native” ATX 3.0 power supply or power supply you buy a modular ATX12VHPWR cable direct, these problems would not exist. Unfortunately it was a premature assumption.
Cases of connectors fused also with direct cables are already known, that is with the MSI MPG A1000G and MSI MEG Ai1300P power supplies. That they are the sources of this brand, but it may not be their fault, because not many ATX 3.0 models have yet been released and MSI was one of the most aggressive brands, its sources came out among the first. Users have posted reports of the two incidents on Facebook and Reddit. It exists there wire, in which, however, 19 users reported overheating and failure of the source connector or adapter and possibly also the port of the GeForce RTX 4090. Of these, two are cases in which a standard cable was used. Since there will be few users with the new source and most will logically use the supplied adapters, that doesn’t mean there is less chance of problems with a direct cable from the source.
In the case of the accident with the MEG Ai1300P (and the MSI GeForce RTX 4090 Gaming X Trio graphics), the user claims to be aware of the reported problems and specifically used a straight cable and tried to bend it as little as possible . However, the connector began to melt – damage appears to have occurred to the pins on one side of the connector, while the other side survived in better condition, so the problem with uneven current load distribution was evidently still present.
It is possible that straight cables (but probably adapters as well) would be safer if they used an angled end, which would avoid the need to bend the cable at the board. Such a well-made ninety degree connector (from which the cable would continue vertically down) should probably be the safest fuse for now. Some of these angled cables or adapters have already appeared on the market, but you will need to purchase them separately. Unfortunately, only straight connector adapters are provided with the cards, which you simply have to fold, because the space between the card and the side of the case typically has a few inches to spare at most.
New recommendation: be careful with full insertion
Corsair’s Jon Gerow (JonnyGuru) also commented on the situation with adapters and cables causing overheating and damage, stating that he personally failed to melt the connector while trying to provoke it (the maximum temperature detected by the thermistor would have been 53 degrees) . But he offered another guess as to what might be behind the uneven resistance and subsequent overheating.
It is said that it is possible that in some cases the user thought that the connector was already well inserted, but in reality it was not fully inserted, causing irregular contact. This is probably quite a risk, because the 12 + 4 pins, at least according to some user reports, fit extremely rigidly, so it could be a problem to push them all the way down after the card is inserted into the card. However, the article appears to have been withdrawn from the website in the meantime (link ends in 404), so we do not know if the finds have been re-evaluated in the meantime.
However, in addition to the recommendation to bend the cables as little as possible (or use a cable or adapter with an angled connector), there is another safety recommendation and something that will be good to check with the graphics: make sure the connection is perfect. all the way when you buy a card with 12 + 4 pins.
So it might be better to insert an adapter or modular cable into the card before inserting it into the PCIe slot, because in this situation you can still easily apply back pressure and visually check that the connector is fully inserted and clicked.
Sources: Tom Hardware (1, 2) Reddit, VideoCardzAndreas Schilling (1, 2)
/nginx/o/2022/11/09/14950647t1h0eb9.jpg?resize=150%2C150&ssl=1 "I wish I was happy all the time. An almost open conversation with comedian Jānis Skuteli")