Multithreaded Architecture: It is clear now that if we give many contexts to multiple threads, then processors with multiple contexts are known as multithreaded systems. These systems are executed in a manner same to multitasking systems. A multithreaded processor will suspend the current context and switch to another. In this way, the processor will be busy most of the time and latency problems will also be optimized. Multithreaded architecture depends on the context switching time between the threads.
The switching time should be very small as compared to latency time.
The processor utilization or its efficiency can be calculated as: U = P / (P + I + S)
P = useful processing time for which processor is busy
I = Idle time when processor is waiting
S = Context switch time used for changing the active thread on the processor.
The purpose of any parallel system is to keep U as high as possible. U will be high if I and S are negligible or very low. The thought of multithreading systems is to decrease I such that S is not increasing. If context-switching time is more when measure to idle time, then the purpose of multithreaded systems is lost.