My case is that these are Location/Position values from the VR Motion Controllers that I want replicated ideally at the VR 90hz frequency.
I could go down in precision to say 1 decimals as that would still give me millimeter moves.
If I assume that the Player will have a max 10 meter side play space, 1000 units, then yes maybe I should be able to use a short style int of 2 bytes each.
But then I have to figure out how to do that in c++, which might not be too hard?
UPDATE: For anyone interested I managed to do this with some small bits of c++. Ideally it should have been possible to just use an int16 per member but it seemed that Unreal does not allow that. So I made a custom struct where each of the vector members was represented by two uint8’s. Then you can google how to turn an int16 into two uint8’s in c code.
So I ended up with a custom struct that gets replicated and two functions to convert a vector into and out of this struct.
It works a treat for my use case and saves a big bunch of data bandwidth. (I think the NetProfiler reports too small size for the struct though)