Think of it like a club with a max capacity of 10 people, where some people have VIP cards. If a person with a VIP card wants to get into the club, the bouncer will kick out one of the people inside that doesn’t have a VIP card to make space for them.

For a more technical explanation:

There are several processors on computers and each can be in use by 1 process at a time. Different processes can get different amounts of time based on their priority (called niceness in Linux) and they’ll be removed from the processor once their time is up until their next share of time.

On a real-time kernel some processes are marked as real-time (certain range of niceness values, can’t remember the exact range). If a process that is real-time says it needs some processor time, a process that isn’t real-time that’s currently running will be immediately ripped off the processor to make room for the real-time process.