Eh, it’s also the connector’s fault. Not saying they’re not to blame for creating junk, but it’s a bunch of companies (pcisig) who created the standard. Even Sapphire’s implementation is flawed.
Doesn’t change the fact that historically balancing the wires on the connector was the job of the GPU. Arguably the connector spec should include who should load balance the wires, it didn’t and afaik it doesn’t, but the established practice was that the GPU takes care of it.
It was Nvidia that designed the original connector and forced it upon the world. PCI has been trying to make it less bad, but it was standardized after it had already been created, not the other way around.
Overcurrent protection on each pin should definitely be mandated by the standard.
But it’s important to keep in mind that Nvidia has 90% market share and can do whatever they want. If PCI standardized something Nvidia didn’t agree with, then there simply would not be any implementations of the standard, and Nvidia cards would use a non-standard connector. It’s that simple.
We successfully shifted the GPUs responsibility from the 3k-4k€ GPU to the 200€ PSU. Fuck Nvidia.
Eh, it’s also the connector’s fault. Not saying they’re not to blame for creating junk, but it’s a bunch of companies (pcisig) who created the standard. Even Sapphire’s implementation is flawed.
Doesn’t change the fact that historically balancing the wires on the connector was the job of the GPU. Arguably the connector spec should include who should load balance the wires, it didn’t and afaik it doesn’t, but the established practice was that the GPU takes care of it.
It was Nvidia that designed the original connector and forced it upon the world. PCI has been trying to make it less bad, but it was standardized after it had already been created, not the other way around.
Overcurrent protection on each pin should definitely be mandated by the standard.
But it’s important to keep in mind that Nvidia has 90% market share and can do whatever they want. If PCI standardized something Nvidia didn’t agree with, then there simply would not be any implementations of the standard, and Nvidia cards would use a non-standard connector. It’s that simple.