Not with the B550 or X570 chipsets, you can't; only Threadripper and Epyc. The RX 5700 is already bandwidth-limited by PCIe3.
So the boards that already exist, don't exist then, both the X570 and B550 versions?
The Asus X570 WS Pro-Ace is wired as follows "Three PCIe 4.0 x16 slots with optimized lane arrangement of 3-way x8/x8/x8 to accelerate an increasingly diverse array of workloads"
This means the top two slots are wired to support 8x PCI-E 4.0 devices, and the bottom slot is using the PCH lanes to make a single 16x slot wired to 8x PCI-E 4.0, with the limitation of 4x 4.0 speeds through the PCH. You also still have the full speed 4x NVMe 4.0 directly from the CPU, and a second M.2 slot that only has 2x 4.0 lanes, or the same speed as 4x 3.0 lanes when using a 4.0 drive, again limited if you are saturating the PCH from the PCI-E 16x slot.
Can you provide some evidence that the RX 5700 is bandwidth limited by PCI-E 3.0 when the RTX 2080 Ti isn't. Seems counter-intuitive that a slower card would require more bandwidth from the system bus.
Except that we have PCIe4 NVME drives and PCIe4 GPUs so you cannot halve those numbers.
The required bandwidth for a GPU is equivalent to 8x PCI-E 4.0 lanes or less until such time that much faster cards become available, and yes we have 4x NVMe PCI-E 4.0 drives, and I have added that in the number I posted if you noticed, I said 4x/8x lanes required for two devices. It is also depending on the speed or use of bifurcation on a PCI-E slot, as you can use a single 16x slot for 4x 4x devices when supported in the BIOS, or an 8x wired slot for 2x 4x devices.
More PCI-E lanes simply are not needed at the level of these CPU, however I would agree that they could offer a half-way house (assuming the I/O die would fit) on the high-end 12/16 core models, with a separate part number and design, if they feel they could justify the redesign, and offer it at a price premium between the TRX40 systems and the 'normal' AM4 boards.