Doug S (foo.delete@this.bar.bar) on February 21, 2019 7:40 am wrote:
> dmcq (dmcq.delete@this.fano.co.uk) on February 20, 2019 3:14 pm wrote:
> > Doug S (foo.delete@this.bar.bar) on February 20, 2019 2:30 pm wrote:
> > > nobody in particular (nobody.delete@this.nowhe.re) on February 20, 2019 7:35 am wrote:
> > > > Write-up at Anandtech
> > > >
> > > > The stated SPECint numbers are superb. I'm looking forward
> > > > to spec2017, but this is a very, very promising start.
> > >
> > >
> > > The ~37 estimated score Anandtech quotes for SPECint2006 is pretty much what the A11
> > > scores. By the time this thing ships, Apple will have the A14 nearly ready. Imagine
> > > if Apple filled a die with those cores and gave it additional memory channels and a
> > > mesh, let alone if they did a design targeted at power budgets above 5 watts.
> > >
> > > If it wasn't for how sad the ARM server story has been up
> > > to now, this thing would be generating a lot of yawns.
> >
> > They're not competing with Apple, they're competing with Intel.
> > The question is where will Intel be by then in servers?
>
>
> Now that AMD is pushing them, and presumably they will finally have 10nm fixed (to some approximation
> of the word "fixed") by then, probably quite a bit ahead of where they are today. And where
> will AMD be then, since they will have server CPUs based on TSMC's 7nm EUV process shipping
> well before this sees the light of day, and 5nm ones on the way for 2021.
>
> The comparison with Apple was merely meant to illustrate that this CPU can only match single core performance
> with a mobile core that shipped 17 months ago. Imagine if Apple was trying. Or what others who might try could
> do if a real market for ARM server CPUs is proven. This thing can only compete with Intel and AMD servers by
> seriously undercutting them on price - even then, how much of the sales price of a typical 2S server is CPU,
> versus DRAM, storage, networking etc.? It might have trouble competing even if the CPU was free...

What you say is true, but what’s interesting here is the uncore IP.
Apple will do its thing, and that’s great; but having this quality of uncore features available to everyone else is a Great Leap Forward, like having a good enough (albeit not the best possible) ARM mobile core available as IP has benefited us all.