This looks like some abomination of the ancient molex drive connection with part of an ATX main connection.
At which point does the GPU get its own dedicated power supply?
At which point does the GPU get its own dedicated power supply?
You think in datacenters they are just stacking up PSUs? No, all of this stuff gets vastly more efficient and reliable if you can build just one big PSU to supply say 48V DC and distribute that everywhere.
The question was more rhetorical than serious.
A GPU these days can often do general purpose compute, but not in a way that meaningfully allows it to compete with the OS for heavily branched or random access workloads since its memory controller and execution model are a poor fit. I'm not sure we'd ever see GPUs fully displace CPUs for that reason.
(See Slide 5): https://papers.put.as/papers/firmware/2014/2014-10_Breakpoin...
A US power outlet is typically rated at 120VAC@15A, that's a 1800W budget if the only thing on the circuit is a PC.
AMD and NVidia have at times competed on power usage, but generally that doesn't sell units to the degree that game support or new features do.
For mobile you start getting traction there because '<vendor>-based laptops have an hour of extra battery life' is meaningful, but gamer laptops are typically designed to be plugged in. The higher power draw typically means more heat and louder fans as well but for some reason noise doesn't seem to be a consideration for most gamers - every gaming laptop I've ever used was unbearably loud.
I'm not sure if any released graphics card has used their own power supplies. I think they all used some sort of internal connection to the standard power supply if the slot can't power it alone.
This is true in many machines already for several years, if you look at the machines the hard to find GPUs are actually in.
I wouldn’t be surprised if that is a market force behind this new PCIe standard.
Its like an acknowledgement of reality is seen as an endorsement of that reality.