Anon (no.delete@this.spam.com) on April 11, 2021 4:40 am wrote:
> juanrga (nomail.delete@this.juanrga.com) on April 11, 2021 4:28 am wrote:
> > And I remark again that N1 core don't need to be clocked at the
> > same frequency than x86 cores because it has an higher IPC.
>
> Have you looked at Zen 3 scores?

Hush, don't burst his bubble. In his world, TX3 is so awesome despite in the real world, the entire team has been let go over 6 months ago. Maybe Marvell has an AI bot that will create them custom cores. Or that namd is such a representative workload despite it needing only around 0.5MB of memory footprint (at least the inputs used in spec) and thus scales linarly in physical cores (but SMT is likely to hurt more than help given the tiny memory footprint). Because cloud workloads look so much like namd, you see. Or even HPC workloads look so much like namd that exascale and super-computing projects in the US and elsewhere are going with Ampere altra N1-based cores. AT even published a per-core perf under load this time around, but some people can't be swayed: https://www.anandtech.com/show/16594/intel-3rd-gen-xeon-scalable-review/8