Some GeForce RTX 5090 fight melty connectors with thermal pads

The power connector joins the Things That Need To Be Cooled On A Graphics Card PCB club

Since GeForce RTX 4000, Nvidia graphics cards (fortunately just the highest models) have been facing issues with overheating and melting 12+4-pin power connectors. Newer models use the 12V-2×6 connector, which is designed to avoid improper attaching, and we recently discussed additional measures from Zotac. It seems that some of the freshly released graphics cards use another protective measure, addressing the root thermal issue.

As noted by Tom’s Hardware (and likely other attentive reviewers and observers), some of the non-reference GeForce RTX 5090 graphics cards reviewed on January 24th (the day after the Founders Edition reference model from Nvidia appeared in reviews) came up with a clever idea: the cooler has been modified to directly assist with the temperature of the hard-pressed connector. It seems like something that, in hindsight, is an obvious improvement, but even obvious ideas can sometimes go unnoticed for a long time.

In the teardown conducted by techPowerUp, it can be seen that cards from Asus and MSI (but not necessarily only those) have a thermal pad placed precisely at the location of the 12+4-pin connector on the PCB. This pad carries some heat from this area to the card’s backplate, which acts as a kind of heatsink. Such pads are often placed under the backplate on the other side of the areas like memory chips or VRMs, as the significant thermal output from these components spreads in all directions and heats up even the opposite side of the PCB beneath them. Therefore, the backplate provides an opportunity to dissipate some of this heat, which in turn reduces the load on the cooler on the main side of the graphics card with the GPU. Using this solution for the power connector as well appears to be a new innovation.

GeForce RTX 5090 Suprim SOC from MSI. One of the cards that adds a thermal pad to the 12+4pin power connector

Unfortunately, the power connector on graphics cards is also positioned in such a way that it lies on the opposite side of the PCB than the backplate. This means that the thermal pad only helps indirectly. The connector must first heat up the PCB before a significant amount of heat can flow to the other side of the PCB. To some extent, however, this should always provide some benefit.

It appears that on the graphics cards in question, the connector might also be cooled directly by the airflow from the heatsink fins, which can blow onto the plastic port (this might not be entirely new, but it’s possible that on some problematic RTX 4090 cards, the issues were aggravated by connector being positioned in a spot that was outside the airflow paths). Unfortunately, the connector is designed in such a way that it is the four signal wires (which do not heat up) that are exposed to the airflow first and foremost, and the problematic power-carrying contacts are located beneath them, closer to the PCB. The plastic housing isn’t best at heat convention, but it is the plastic housing itself that obviously needs to be cooled down to prevent the melting damage, even if excessive temperature of the metal contacts inside also needs to be reduced (because it can damage to the cable insulation from the overheated wires).

ROG Astral GeForce RTX 5090 from Asusu. One of the cards that adds a thermal pad to the 12+4pin power connector

This cooling design could improve the situation with connector overheating. The advantage is that it actively reduces temperature, so it has some effect in all situations that lead to increased temperatures, including cases where deformation and uneven resistance occurs due to twisted wires, defects, or poor-quality connectors.

Until now, most measures against connector melting have focused on preventing improper insertion of the connector, including the revision of the 12V-2×6 connector itself. Whether these measures fully address the risk depends on whether improper insertion is truly the only possible source of problems, as Nvidia claims. However, Nvidia, being essentially the originator of the entire issue, is not in a completely neutral and unbiased position. It’s hard to be sure the company is not trying to downplay the fundamental fragility of the connector and the risk that it could fail even when fully inserted.

Anyway, it would be nice if this practice with a thermal pad under the connector and other cooler modifications to improve the temperatures of this part of the cards become standard. This approach doesn’t seem to be used to cool the connector on the RTX 5090 Founders Edition graphics card, where it looks like the critical location doesn’t exactly allow air from the heatsink to blow onto the connector due to how the fans are positioned and the impenetrable cooler baseplate.

Source: Tom’s Hardware

English translation and edit by Jozef Dudáš


  •  
  •  
  •  
Flattr this!

Leave a Reply

Your email address will not be published. Required fields are marked *