That is the right way to do it I guess.
Maybe bigger values (4MB/1MB) would be needed for faster 2.5/5/10G ports and WLAN 5GHz 160MHz - 2400Mbit/s.
But testing is the right way and it's good that we now know the reason for the speed limitation cap that we had.
For my usual usage currently I don't see any memory penalty but of course if I see higher memory usage I'll report it.
Regarding the memory usage this @KONG's post maybe usefull too.