Generally, there’s a small difference in speeds:

PCIe 2.0 x 4 > 2.000 GB/s PCIe 3.0 x 2 > 1.969 GB/s

But we also have to consider the suggested ARM CPU does PCIe 2.1 and we’ve to add the this detail:

PCIe 2.1 provides higher performance than the PCIe 2.0 by facilitating a transparent upgrade from a 32-bit data path to a 64-bit data path at 33MHZ and 66MHz.

I shouldn’t also have a large impact but maybe we should think about it a bit more.

Anyways I do believe this really depends on your use case, if you plan to bifurcate it or not and what devices you’re going to have on the other end. For instance for a NAS I would prefer the PCIe 2.1 x 4 as you could have more SATA controllers with their own lanes instead of sharing lanes in PCIe 3.0 using a MUX.

Conclusion: your mileage may vary depending on use case. But I was expecting to have more PCI lanes exposed be it via more m.2 slots or other solution. I guess that when a CPU comes with everything baked in and the board maker “only has” to run wires around better do it properly and expose everything. Why not all SATAs for instance?