Last weekend, we focused on the issue of 12VHPWR and 12V-2×6 connectors on Nvidia graphics cards, particularly what Der8auer’s findings revealed about their reliability and safety. Beyond the problems with cards like the GeForce RTX 4090 and 5090, this issue has another manifestation: while Nvidia claims the problem no longer exists, other hardware manufacturers have started coming up with their own ideas to address or prevent it.
We’ve actually seen some of ideas of this sort before—like the cable insertion correctness indicator from Zotac (though, like colored connectors, this only ensures proper cable insertion, which isn’t 100% protection against cable failure), or the use of thermal pads in cards to slightly cool the connector if it starts overheating due to contact resistance.
Thermistor on ASRock and FSP power supplies enables overheating detection
ASRock has introduced a more active protection mechanism in its PG-1600G power supplies compared to just employing the colored connectors that many manufacturers have started using (as pointed out by Aris Mpitziopoulos from the YouTube channel Hardware Busters). Or rather, this might be an idea from FSP Group, which manufactures these power supplies for ASRock.
The PSU puts two additional pins next to its modular connectors for 12+4-pin cables, and the included 12+4-pin cables have two extra signal wires that connect to the PSU. These wires carry a signal from a thermistor, a temperature sensor that ASRock has integrated into the cable. The sensor itself is located near the connector that plugs into the graphics card.
Thanks to the thermistor, the PSU has information about the temperature near the connector on the card, and if it gets too high, the PSU’s electronics can stop delivering power to the GPU. This shuts down the GPU and potentially prevents damage. According to Aris Mpitziopoulos’s test, the PSU cuts off power when the connector reaches around 110 degrees Celsius. However, this might be somewhat influenced by the methodology used—the contacts and wires weren’t directly heated, but rather a heat gun was applied to the exterior.
The use of this sensor is ultimately following comparable goals to Asus’s approach, which equipped the individual 12V wires on the GeForce RTX 5090 Astral graphics cards with shunt resistors to monitor the current passing through each contact. However, in this case, ASRock is tackling the issue from a different angle. Asus can alert the user in case of abnormal readings, while ASRock’s solution aims to detect overheating directly and shut down the card automatically (without relying on a manufacturer’s app, as with Asus).
The question is how effective this cable is and whether cutting off power at 110 degrees Celsius is too late, or if it perhaps would be better to shut down the GPU earlier to prevent damage. Additionally, it seems the thermistor is only on the end of the cable that plugs into the graphics card. However, current cases of burnt connectors on the GeForce RTX 5090 and RTX 4090 show damage on both ends of the connector, or even just on the PSU side. Protection would also be useful there, especially if the other end has a 12+4-pin connector, as with this ASRock cable. It’s possible, though, that the PSU has an internal thermistor directly on its internal PCB, where it can measure the temperature on the plastic part of the plug from within the PSU enclosure.
Should cables have fuses or similar protection?
There have also been proposals that cables could integrate fuses or other methods of monitoring current (such as shunt resistors) across the wires to prevent scenarios where the load becomes unbalanced and causes most of the current to flow through just a few or even one wire, burning the contacts and melting the connectors. The idea is that the cable itself could detect such a problem, since the graphics card doesn’t (and it’s not entirely appropriate for the PSU to do so).
Implementing this isn’t trivial, as it’s not just about detecting an overload on an individual wire but also about cutting off power to all wires in such a case. So, it’s not just a matter of adding fuses somewhere on the cable wires—it would require a more complex system. Such a cable would inevitably be more expensive than the standard ones, unfortunately.
Will we have cooled cables?
The company EZDIY-Fab has come up with a simpler solution—cable connectors that are cooled, albeit passively (let’s hope active cooling with fans isn’t the next stage of the entire 12V-2×6 cable “SNAFU”). EZDIY-Fab offers 90° and 180° adapters for power cables, which feature an aluminum cover designed to act as a passive cooler. It’s not finned, but it has internal vertical channels for airflow (though the question is how much air will actually flow through them).
Something similar is offered by CableMod, a company that manufactures modular cables for graphics cards and has previously learned about from the weaknesses of the 12+4-pin connector standard the hard way. Its current adapters use PCBs with a higher copper content and additional copper foil for heat conduction. An aluminum cover is then attached via a thermal pad, serving as a cooler.
It’s worth noting that a cooled adapter might not always help much, as the connector of the cable plugged into it could still overload and isn’t cooled. Generally, with these sensitive connectors, every additional connection is a potential extra risk. In fact, limiting the number of 12+4-pin connectors that can fail is one reason why it might now be safer to use adapters from multiple 8-pin connectors instead of direct 12VHPWR/12V-2×6 cables.
However, the idea of cooling the cable connector could also be applied to direct cables and even to the graphics cards themselves. Given the current issues, manufacturers might want to focus on ensuring the connector is exposed to airflow from the card’s fans or otherwise improving heat dissipation.
Bandaids instead of solving the problem?
Of course, the question is whether it’s appropriate to complicate cables in this way instead of acknowledging that the connector itself is the problem and replacing it with a better alternative. Personally, I’d say that all these attempts are trying to mitigate an issue that didn’t need to—and never should have—happened. However, abandoning the inadequate 12VHPWR/12V-2×6 connector standard is something that depends on Nvidia, which has so far remained silent on the issue (the company hasn’t even released a statement acknowledging the problem).
Given that Nvidia is currently the chip company with the highest revenue and market capitalization (with expectations of further growth), it unfortunately won’t be easy to push them toward a solution if they don’t choose not to act. But that shouldn’t discourage efforts to do so. This includes, for example, not buying problematic graphics cards with 12+4-pin connectors and power consumption over 300 W, and instead opting for lower-tier models that are safer due to their lower power consumption.
Sources: Hardware Busters, ASRock, Tom’s Hardware (1, 2), EZDIY-Fab, CableMod
English translation and edit by Jozef Dudáš
⠀
