CRamseyer :
I was talking about the surface area of the M.3 compared to the M.2. I need to see what edits were made. Also, you can fit more M.3 drives in a 1u server than 2.5" drives. I think SuperMicro has a 1u with 10 2.5" drive bays. There are 34 in the server shown with M.3.
Gotcha - I see what you mean.
I could argue that if you had a full-size (length & width) 2.5" board with maxed-out amount of memory modules (not including heat-removal mechanisms) instead of the current way wherein only half (or less) of the interior space is used, then I would think you should be able to get *more* storage. You would have *one* controller with more memory modules than the M.3/NGSFF form factor where the single controller only has the memory modules that can fit onto its small surface area.
For example, I have one M.3 module with 4 memory chips. If I lay two down side-by-side, let's just say that equals the width of a 2.5" board. So I would have two controllers & 8 memory chips total. However, on the 2.5" board, I only need the one controller, and I can then stuff more memory chips onto the board than just 8 chips -maybe 10-12 total. So physically, I have more memory chips on the same surface area (i.e. more storage space).
I know the consideration was lining up the M.3/NGSFF drives vertically in a 1U server. Maybe the same could be done if you stacked 2.5"-width boards (sans case), and maybe not. However, my main thought was that physically, a 2.5" board can offer more space for memory modules than a M.3 board.