Let's say I have a server cpu with 18 cores, with hyperthreading on, which means I can see 36 cpus in htop.
To fully utilise the CPU and not impact single-thread performance, should I be aiming for all 36 "cores" to run at 100%, and the HT cores will just do less work and still report 100%, or would having that mean the "full" cores are already getting interrupted by the task on its "HT core" and thus doing less single-threaded work?
I'm aware that there are a lot of variables that affect HT performance, I just want to know what cpu meters mean when dealing with HT.