Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It's not NUMA since the distance from all the cores to main memory is the same (uniform).


Although Threadripper 2990 is not uniform to main memory, which adds a wrinkle

So now there’s at least 2 levels of Non Uniform access to manage by the Bios/OS


Don't forget the CCX level, too.

The full list of varying latencies is something like: SMT, inter-core, inter-CCX, inter-die, inter-socket. And even that misses a few subtleties.


Oh hrmm, my mistake. This is a welcome change from the older one.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: