When will “they” invent a better bus to go between the system memory used by the CPU and the GPU? It seems like that is a pretty major bottleneck. It took a wh

When will “they” invent a better bus to go between the system memory used by the CPU and the GPU? It seems like that is a pretty major bottleneck. It took a while before SSD’s stopped using SATA despite SATA being a huge bottleneck. Now it uses something else whose name I cannot recall. Surely there is work to do something like that for the GPU. Because while it’s been a long while since I’ve built a computer I do know that the video card has always been the peripheral that pushed the limits of whatever interconnect bus existed at the time. There was all kinds of hacks for it like AGP, Video Local Bus, and then even PCI express to a large degree.