When PCI-SIG introduced the 12VHPWR power connector as a replacement for the 6- and 8-pin PCIe power connectors, it created a wave of controversy. There were enough cases of melting GPUs, PSUs, and cables to set people on edge. Amidst this controversy, [JayzTwoCents] decided to do some scientific experimentation, Mythbusters-style, specifically: do these 12VHPWR (or the 12V-2×6 successor) wear out upon hitting the often cited 30 mating cycles? If this is the case, it could explain why people see wildly different current loads between the wires in their GPU power cables. Perhaps reviewers and hardware enthusiasts should replace their GPU power cables more often.
Like many Mythbuster experiments, the outcome is less than clear, as can be observed in the below graph from one data set. Even after 100 mating cycles, there was no observable change to the contact resistance. One caveat: this was only performed on the GPU side of the connector. The first cable tested was a newer connector type that uses a single-split leaf spring design. Initially, most of the 12VHPWR connectors had a double- or triple-dimple design to contact the pin, so [Jayz] tested one of these, too.
The amazing thing with the 2022-era cable that got pulled new out of packaging and tested was that it looked terrible under the microscope in terms of tolerances and provided a very uneven load, but it got better over time and also lasted 100 cycles. However, it must be said that ‘lasted’ is a big word here, as the retention tab wore off by this point, and the connector was ready to fall out with a light breeze.
Perhaps the ‘mating cycles’ specification is more about the connector as a whole, as well as how the connector is (ab)used, at which point good (long-term) contact is no longer assured. Along with the different types of Molex Mini- and Micro-Fit style connectors, it’s worth keeping an eye on with more applications than just GPUs.
We have certainly seen some burned connectors. Particularly in 3D printers.
Why reinvent the wheel when there are welding cables with connectors good for at least 250 amps.
Because some people will then “weld” their GPUs by connecting them to their welder instead of “new” PSU…. ;-)
Companies protect their products with their brandnames, if you use their property, the sue you. That is the stupid side of our world economy.
Did a quick search for “high power pcb connector” found multiple similar designs for the (automotive) industry:
https://vehicle-electronics.biz/content/phoenix-connectors-rise-tti
It would be rare for a connector to break at exactly the rated cycle count, or even 4x that. It’s that some units will be worse off to start with, and then the problems usually only increase over use. If the manufacturer hasn’t just pulled the rating out of thin air, it is likely derived from statistics over a large sample of connectors, some of which broke earlier and some later.
There are a surprising number of common connectors out there specified with very limited insertion cycle ratings, SATA connectors, car ECUs etc. tend to have very low figures.
Basically because they’re designed to be installed and used until they fail, which the vast majority of items in the field are.
I think high current connector work only well and reliable if they made by a good experienced company and only if both parts made by the same company! And of course only if the people that use them have some experienced in look and feel when connecting.
For this reason I think this kind of connector are a wrong choice for the combination of current, user and a world where every company try to make anything as cheap as possibly.
Perhaps they should rise to voltage to 42V. There a already POL-Converter on the cards.