We can add another Radeon 9070 XT casualty to the list of graphics cards damaged by a burnt 12V-2×6 power connector. AMD’s partners began adopting the 2nd-gen 12VHPWR connector with the Radeon RX 7000 series, and while it took a bit of time, stories of melted connectors have been popping up with increased frequency. This time around, it’s barely been 3 weeks since the last, which was also a Sapphire Nitro+ 9070 XT, and in both instances, there seems to be another common denominator.
There’s a long and torrid history for owners of newer NVIDIA GPUs regarding the 12VHPWR/12V-2×6 connector, but there are also a myriad of reasons for burnt connectors. They range from user error to poor-quality after-market cables, a lack of load-balancing technology, and using improper cables or PSUs. Regarding the latest instances involving Sapphire’s GPUs, both (1, 2) users connected their GPUs via the included 3x 8-pin adapter to non-ATX 3.1 PSUs. As these GPUs draw significantly less power than the maximum 600W, and feature blue-colored connectors to assist in visual confirmation that they are fully inserted, it would seem that there could be a load-balancing issue when using the adapters with non-ATX 3.1 power supplies.
Technically, this makes the third 9070 XT to experience this issue since August, but the first, an ASRock Taichi OC 9070 XT, was connected unconventionally to an 850W PSU using the 3x 8-pin adapter, but with only two of the 8-pin connectors plugged directly into the PSU and the third via another adapter. Clearly not an ideal or recommended method, but as mentioned by that user, the PSU wattage did meet the manufacturer’s recommendation. Perhaps the biggest takeaway, for right now anyway, is that owners of any AMD Radeon 9070 XT partner cards that utilize the 12V-2×6 connector may want to consider not using the 3-pin adapter and upgrade to a newer ATX 3.1 PSU and use its supplied 12V-2×6 cable instead. However, as the many reports for NVIDIA GPUs have shown, this is not a cure-all either, but things are not looking good for the 3x 8-pin adapters right now for the Radeon RX 9070 XT series.
Discussion (11 replies)
Join Discussion →It's just a trash connector yo.
The problem doesn't look common, but still far too prevalent for my tastes.
Wow I would have never er expected a card with this lower power draw to even be capable of this sort of disruption.
The 2x6 connector is just a poor choice. The contacts are small and often not fully seated in the plastic. There are not many choices that inspire confidence if you own an older PSU. I havent had any issues yet in testing many current gen cards but thats no guarantee. I do use an ATX 3.1 in my testing platform and my personal PC, its the best you can do.
So far so good with all the GPUs I've owned with these connectors (2x 3090 Ti's (used 4x 8-pin adapters at first for these), 4080 SUPER, 4090, and now 2x 5090's) but I also made the upgrade to ATX 3.1 back when I got a 4090. We all know there can be other contributing factors but I wanted to ensure I did my part for the best I could do.
Scrap the connector, go back to the drawing board. This design should have never passed PCI SIG.
This may be true, but if you google melted 8 pin connectors you will also find those so that's not the solution either then. Anything will break somehow eventually, if the amount is too high they will change it sooner rather than later the problem is getting to know the amount of defects vs amount sold.
Given current high end GPU's power draw something a lot more robust is needed, preferably soomething less prone to user error.
They need to just start shipping those high end GPUs with mini nuclear reactors in them. No more melting of connectors then.
...notice I didn't say no more melting cleanly. LOL
GPUs need to move to a mezzanine connector which can deliver enough power through the motherboard, and securely mount a cooler that integrates with the case. The ATX card format was never designed for this much heat generation or power consumption, it was meant for cards that look like this:
This, or really, if you think about it - a current PC design is focused around the CPU. But the CPU no longer consumes most of the power, and in many cases, doesn't do most of the processing.
On these high end builds really they should be built up around the GPU - the motherboard should be the GPU board itself, and everything else should just plug into it. The CPU should be moved to a daughter card...
I agree, we're at a point where a complete redesign needs to be worked on. We've seen for several gens now how the GPU is often bigger than the rest of the build. That alone is a clear indicator of a shift in processing power and needs.
Agreed. It was an issue (albeit a small one) with the 4090 so a redesign is overdue.