The benefits of PCIE 4 come into play with stuff like offloading heavy computation to an FPGA/ASIC type device (like the Acorn)...since you can now offload heavy compute tasks twice as fast between the GPU and secondary device through the PCIE bus. So for heavy memory bound algorithms like most of the GPU centric ones you can have the GPU do what its good at and handle the large memory intensive parts of the algorithm, and then offload the compute heavy rounds to a FPGA/ASIC device. This will significantly reduce the power consumption of the GPU and give a hash rate boost, and the speed ups will be compounded due to the bandwidth increases between the GPU and secondary device over PCIE 4.