As others have stated, thread switching overhead is much higher than even the cost of a main memory fetch, so we can't afford to switch a thread on every pipeli

As others have stated, thread switching overhead is much higher than even the cost of a main memory fetch, so we can't afford to switch a thread on every pipeline stall. Numbers Every Programmer Should Know: https://people.eecs.berkeley.edu/~rcs/research/interactive_latency.html https://people.eecs.berkeley.edu/~rcs/research/interactive_l... (Although I see that this does not include thread switching latency, possibly because it depends on the OS and thread type. I'll leave the link because it's generally useful.)