The interrupt needs to become active as often as required. If your task asks for repetition in 1 µs intervalls then this is as it is.
How much the controler can work outside the interrupt routine depends on the length of the interrupt routine (interrupt routines always should be as short as possible) and on the speed of the controller:
- A controller operated from a 1 MHz clock will not be able to do anything, not even within the interrupt routine as the interrupts come at the same rate as the clock frequency.
- A controller operated from a 50 MHz clock can at least perform 49 operations between 2 interrupts (assuming 1 operation per clock cycle).
- A controller operatde from a 50 MHz clock with pipelined instruction processing may be able to perform many more operations.
You see: your question cannot be answered in a completely general and generic way.
On the other hand, inspect your requirements carefully: do you really need one interrupt every 1 µs? Would it be possible to outsource the operation from the interrupt routine into external hardware which would pre-process the data and interrupt the controller less often?
An example of the latter is digital filtering of an input signal. You can do the filtering in real time using the controller but you may have a heavy load on the controller. You could also use a hardwired external filter, e.g. in an FPGA, and use the controler to process the already filtered signal at a much lower data rate.
