By: Chester (lamchester.delete@this.gmail.com),
Room: Moderated Discussions
> > > Here's the actual picture:
> > >
> > > Absolute scores from AT:
> > > - N1: Probably 37
> > > - i9-9900K: 59.45
> > > - Ryzen 3900X: 57.25
> > > - Ryzen 2900X: 41.89
> >
> > Those are single-thread scores.
And single thread scores matter. That's the advantage of SMT: when your workload only occupies less than half the available threads, per-thread perf gets better.
> > At 2.6GHz N1 it is able to math a higher clocked Zen2 with SMT.
>
> it is able to match a higher clocked Zen2 with SMT.
No, unless you compare two N1 cores with one Zen2 SMT core.
Don't get me wrong, N1 is a competent core. It's probably good enough for a variety of server workloads. And in some cases it's better to have a lot of smaller cores than fewer faster/beefier ones. But N1 isn't some miracle uarch.
> > >
> > > Absolute scores from AT:
> > > - N1: Probably 37
> > > - i9-9900K: 59.45
> > > - Ryzen 3900X: 57.25
> > > - Ryzen 2900X: 41.89
> >
> > Those are single-thread scores.
And single thread scores matter. That's the advantage of SMT: when your workload only occupies less than half the available threads, per-thread perf gets better.
> > At 2.6GHz N1 it is able to math a higher clocked Zen2 with SMT.
>
> it is able to match a higher clocked Zen2 with SMT.
No, unless you compare two N1 cores with one Zen2 SMT core.
Don't get me wrong, N1 is a competent core. It's probably good enough for a variety of server workloads. And in some cases it's better to have a lot of smaller cores than fewer faster/beefier ones. But N1 isn't some miracle uarch.
