Here's an illustration showing corrosion build-up on plated material after the plating is worn through
Good illustration! I now think a major part of the problem could actually be corrosion, not due to cable quality or age but due to the design of 12VHPWR and a (very) limited number of reseatings, much smaller than 30x.
Hypothesis:
With 12VHPWR/12V-6x2 you will likely end up at some point in it's lifetime with corrosion inside the cable due to the contact mating force for the 12VHPWR standard being too high. At least for this kind of connector. When you start to reseat the cable (you will most likely do that at some point in its lifetime), the following script occurs. Of course, my speculation.
1. A few unlucky reseatings in combination with the high mating force are enough to start grinding and eating away the plating metals on the tiny wires
2. Once the exposed copper is out, corrosion kicks in
3. Corrosion then eventually leads to increased contact resistance
4. Which then results in unbalanced amps on other wires as the GPU is 'dumb' and draws the power anyway
5. Eventually one of the six power wires/pins gets to hot and the area around it starts to melt (GPU or PSU side)
6. Bob's still your uncle
Actually, I think (another speculation)
many people with a 12VHPWR GPU (any class) suffer from load balance problems by design -
but this likely goes unnoticed and unpunished most of the time as the load imbalance per wire/pin is probably still within spec. Or the corroded wire points to a sense or ground pin. The 4090 with it's higher power draw was the first to show some problems here and there (the melting issues). The 5090 only takes it up a notch with the increased power draw - so it may be even more vulnerable to exceed the specs.
Also, I don't believe you can rely on the 30 mating cycles of 12VHPWR as a guaranteed cycle count. It can work sometimes, and sometimes it won't. The cycle count should be much smaller with this kind of plug, this kind of mating force, this kind of tiny AWG16 toy wires and this kind of power draw. It's not the fault of the cables, it's a design fault.
Still, we don't know, if it's a full blown issue with the 5090, as almost nobody owns those cards or can buy one.
The user error issue:
Of course, you can mitigate the risk by using the Nvidia adapter since it’s a
new piece of cable and connector. So, just like with the 4090, novice users, Nvidia loyalists and people without load balance problems will flood the forums with the usual
“user error” mantra - blaming people with 12VHPWR issues for not being skilled enough or not RTFM-ing or not using the Nvidia adapter. We already died on that hill with the 4090. One can already see some of those people in this thread (e.g. last 4 pages, quality and value of debate are rapidly decreasing in result). They don’t see the real flaw in this connector design, and they’re incredibly useful to Nvidia in pushing the narrative that this fragile standard is somehow sufficient for a flagship GPU power draw.
I made two 4090 builds - I am typing this on one of them. I am a green guy. I want the 5090 to work (I also want to buy one when the fog clears up). But we all need to keep up the pressure - because we simply need a better GPU connector and better cables in the next gen. Hopefully Nvidia will ditch 12VHPWR again
