As we learned, when several CPUs are installed in a computer, programs can create several threads, each of which will run, possibly, on its own CPU. If we divide some work in a relatively even manner across the threads, we decrease the time it takes for the program to run, and thereby make it faster.
For example, if a certain program contains a loop of code, and each of the loop iterations is independent from other iterations, we can divide the work in the loop among several threads to speed the loop's execution.
If the computer contains, say, 4 CPUs, the program could spawn 4 threads (besides the main thread,) each of which will be assigned to a part of the job (in our example, loop,) thereby speeding the job's time close to 4 times. Why close to 4 and not exactly 4 times faster? Because the operating system spends some extra time on the creation and management of each thread (a.k.a., "overhead",) which reduces the overall speedup.
A pretty simple math formula that calculates the maximum possible speedup is called Amdahl's Law.