The problem is that on top of the pins occasionally not making good contact on these new connectors, Nvidia has been cheaping out on how power is delivered to the card.

They used to have three shunt resistors that the card could use to measure voltage drop. That meant that the six power pins were split into pairs and if any pair did make contact the card could detect it and prevent the card from powering up.

There could be a single pin in each of those pairs not making contact meaning that the remaining pins are being forced to handle double their rated power. It is unlikely that you would lose one pin on each pair so that is an unlikely worst case, but a single pin in a single pair failing could be fairly common.

But on the 40 series they dropped to two shunt resistors. So instead of three pairs, they can only monitor 2x bundles of three wires. Meaning the card can only detect that the plug isn't plugged in correctly if all three wires in the same bundle are disconnected.

You could theoretically have only two out of six power pins plugged in and the card would think everything is fine. Each of those two remaining pins being forced to handle three times their normal current.

And on the 5090 FE they dropped down to one shunt resistor... So five of the six pins can be disconnected and the card thinks everything is fine, forcing six times the current down a single wire.

https://www.youtube.com/watch?v=kb5YzMoVQyw

So the point of these fused cables is to work around a lack of power monitoring on the card itself with cables that destroy themselves instead of melting the connector on your $2000 GPU.

Except the whole point of 12VHBRN over just 4x 8-Pin is gone - looks. Though, if you'd go after NVidia and only use 2nd party cables (so only the adapter), it would look awful too.