juanrga (nospam.delete@this.juanrga.com) on August 16, 2014 12:45 pm wrote:
> David Kanter (dkanter.delete@this.realworldtech.com) on August 16, 2014 11:27 am wrote:
> > > > I happen to know the differences between those two designs. I'm not really sure it's
> > > > going to translate into a significant performance delta. My guess is maybe 10%.
> > >
> > >
> > > Care to explain how you got the 10% figure? It looks a bit low for me. Thanks
> >
> > My expertise in computer architecture and knowing the differences between the two designs.
> >
> > > > > Another key is that Intel process advantage will be reduced.
> > > > > Those server-class ARM SoCs that I >mentioned will
> > > > > be made on 14/16 nm FinFET. Broadwell-EP and Skylake-EP on 14nm FinFET will not have a full node >advantage.
> > > >
> > > > Actually they will. The foundry 16nm process will have substantially worse density. The minimum metal pitch
> > > > for the foundries is 64nm, vs. 52nm for Intel. That's about a 20% difference, which is quite significant.
> > >
> > >
> > > And a ~50% density advantage is not "a full node advantage" as I mentioned just
> > > above. Intel _traditional_ process advantage has vanished, as others agree,
> > >
> > > http://www.electronicsweekly.com/news/components/microprocessors-and-dsps/whats-new-14nm-processors-intel-2014-08/
> >
> > Anyone who believes that doesn't know what is going on.
> >
> > If you look at Intel's 22nm high volume production vs. TSMC 20nm, there's about a two year gap (or more).
> >
> > Remember that what's important is comparing when Intel ships products to customers and when
> > TSMC customers ship comparable complexity products to customers, in similar volumes. TSMC
> > risk production (i.e., hey we got a good die per wafer!!!!) is not comparable to Intel production
> > (i.e., the factory is barfing out hundreds of thousands of good dice).
> >
> > Intel launched IVB in 2Q12.
> >
> > Apple will launch the new SoC in 3Q14.
> >
> > That's basically a two year gap.
> >
> > There's no indication that gap will narrow.
>
> There is evidence that it is not longer two years, see below.
>
> > > > I'm also rather skeptical that we'll see 16nm FinFET in production before the end of 2015.
> > > > TSMC won't be in high volume on 20nm till 3Q when the new iPhone comes out. I expect that
> > > > 16nm FinFET will take at least a year, and quite possibly more to hit high volume.
> > >
> > >
> > > TSMC claims _10nm_ risk production for late 2015. 16nm is being produced now (check above link). ARM server
> > > 16nm parts will be available during 2015--2016. Broadwell has been delayed again to late 2015.
> >
> > 1. Get your facts correct: Broadwell parts will ship this year.
>
> Only the low-volume parts will do. Broadwell-K, -E, and -EP are delayed to Q4 2015.

BDW-EP wasn't delayed. HSW-EP is coming out in Q3 or Q4, and BDW-EP will be a year later. That's not a delay, that's on-time.

I agree that certain mobile and desktop parts were delayed.

> > 2. TSMC may be making 16nm wafers, but I don't care about that (hint: Intel is making
> > 10nm wafers now...). I care when TSMC customers are shipping products in high volume
> > to customers (and I don't count FPGAs because they are very different).
> >
>
> TSMC 16nm is on risk production now. Products are expected in 2015--2016. Thus
> ARM 16nm server products will be competing against Intel 14nm Server products.

If you have a subscription to MPR, you should be able to see one of my articles on proces technology. Let me put together an outline for you about TSMC 20nm:

1Q13: Risk production
3Q13: FPGA production
1Q14: Production
2Q14: First 20nm products sold (TSMC modem in very limited supply to Korea)
3Q14: Predicted high volume for Apple SoC

So judging by history, it will take 6Q for TSMC to get from risk production to actual production. Now I can believe that it will be a bit shorter since they are using the same metal stack, so call it 4Q. That means no high volume products on 16nm till 3Q15...and to be honest, it probably won't be server products.

Server products have much longer validation cycles, so I am skeptical that any 16nm server products will be sold prior to 2016.

David