I don’t think the PCI4 speeds are a difference maker, at least for the GPU. A gen4 framestore is amazing and finally lets me have smooth playback on 10-bit 8K x 8K stereo footage. The PCI link speed, along with bus width, determine the overall rate at which the GPU will have access to resources. AFAIK, this is only really relevant in the most memory intensive applications. Applications that operate very quickly on huge sets of data. It has to be both of these things for the link speed to be a bottleneck. In any other case the link speed barely makes a difference between gen3 and gen4.
Keep in mind, the bandwidth is doubled between the two gens.
Fortunately, this is an easy claim to test: One can force the link speed of a particular PCIe slot to be gen3 or gen4 in bios. After confirming in nvidia-settings that I was running my 3090 in gen3 mode, I ran the Flame 4K benchmark again and found it be 2 seconds slower (4:52 vs 4:54). That’s margin of error.
This would probably yield a bigger difference if one were to use a gen4 card @ gen3 with a lot less VRAM (like 6 - 8GB) when working on 4K or higher plates…but that limits the possible cards down to a 3050-- a card you should definitely not be using for Flame unless you really have to.

