The main flaw I notice in the design of this board, as a BPI-R4 user, is the failure to expose all possible ports/connectors for the user. In the BPI-R4 PRO, this issue is less pronounced, and there are even spots on the board for an M.2 Key E socket.
For example, a major oversight is not exposing the last 2.5GB RJ45 port from the switch (which supports up to five ports, even as pin headers). Second issue is lack of support of tacho/RPM in fans, and common PWM line.
There are a total of 6 PCIe lanes: 2 lanes per PCIe slot, one for the M.2 Key M connector, and one for the Key B connector (in BPI-R4, close to similar situation in PRO). The PCIe slot also has USB 2.0 routed from the hub (just like the M.2 Key B port).
In my opinion, it would be ideal to have an expansion board that converts the dual PCIe connector into 4 M.2 Key M slots plus exposed USB 2.0 ports (2x). The question of power supply sufficiency would need to be considered—perhaps switching to 12V and using step-down converters would be necessary.
I’m not an electronics specialist, but if it’s possible to combine 4 PCIe lanes into a single M.2 Key M slot, the board should have 4 M.2 Key M slots. If only one NVMe drive (or any other device) is inserted, it would use all 4 lanes. If a second drive is inserted, each would get 2 lanes. If three drives are inserted, one would get 2 lanes and the other two would get 1 lane each. If four drives are inserted, each would get 1 PCIe lane.
If PCIe lanes can only be grouped per PCIe slot, the board would have two M.2 Key M slots.
What stands out most to me is the closed-minded vision that everyone will use this board only as GSM modems, when it could be used in so many other ways. It’s a shame that these boards aren’t designed with more versatility in mind.
I’m planning to order the design of such a board somewhere—hopefully, it will be financially reasonable (I’ll probably do it in China).