Just to explain the two numbers:
Concurrent connections refer to the number of connections where state can be maintained and are expressed in millions.
This is mostly a function of available memory.
Connections per second is a function of what the hardware is able to actually pass traffic on.
If you exceed the threshold, packets will get delayed and/or dropped at the hardware level before they are processed in software.
Note that not every connection passes packets every second.
That said, these numbers only apply under ideal conditions.