By: juanrga (nospam.delete@this.juanrga.com), August 18, 2014 2:39 pm
Room: Moderated Discussions
David Kanter (dkanter.delete@this.realworldtech.com) on August 17, 2014 6:42 pm wrote:
> juanrga (nospam.delete@this.juanrga.com) on August 17, 2014 4:59 pm wrote:
> > David Kanter (dkanter.delete@this.realworldtech.com) on August 17, 2014 10:23 am wrote:
> >
> > > > And about total power consumption and efficiency:
> > > >
> > > >
> > > >
> > > > Therein they are mentioning the advantage of SoC vs a CPU. As I said before, the 80W are for
> > > > the whole SoC. The 95W are only for Xeon CPU, adds the TDP of rest of components to the
> > > > Intel platform and you will need up to double power to do the same work than ARM SoC.
> > >
> > > Your calculations are wrong, but your point is good.
> > >
> > > Intel does not (currently) integrate networking, and that leaves an opportunity for a competitor to offer a
> > > different system architecture that is differentiated. That
> > > being said, it only costs about 10-20W to add 4x10G
> > > ethernet MACs. I'm not sure about the actual $ cost, but Intel would simply need to lower their prices.
> > >
> >
> > Add the south Bridge and now consider the performance gap, 350 vs 320, and you will obtain that you
> > will need something close to twice more power to do the same work than the ARM SoC. I got 89%.
>
> Let's compare calculations here.
>
> I've got:
>
> ThunderX: 80W @ 350 SPECint_rate
> Xeon E5-2650L: 159W @ 546 SPECint_rate, including 2x70W CPUs + IOH + 7W for a dual
> 40G ethernet controller, note that you can fit upto 5 NICs for each server CPU.
>
> However, Thunder will probably be compared against HSW-EP or BDW-EP,
> which should be about 30-50% better performance/watt than IVB-EP.
>
> David
We were discussing the chips in the table 1 of the Microprocessor Report article about ThunderX. My above 89% is for the model of Xeon listed in that table. Now you change the model...
I will assume that your computations for the new Xeon model are right. Even assuming that, the TunderX is about 30% more efficient.
Yes, ThunderX will be competing against Broadwell-EP. Taking your best value of 50% improvement. This means that an expensive Broadwell-EP Xeon on 14nm FinFET will be only slightly better in performance per watt than a cheap ARM SoC made on 28nm planar.
But I already said you that ThunderX is not where the real battle will be. The real battle will involve the X-Gene v3, Vulcan, and K12, all them will be made on 14/16nm FINFET.
> juanrga (nospam.delete@this.juanrga.com) on August 17, 2014 4:59 pm wrote:
> > David Kanter (dkanter.delete@this.realworldtech.com) on August 17, 2014 10:23 am wrote:
> >
> > > > And about total power consumption and efficiency:
> > > >
> > > >
Compared with Xeon, ThunderX could deliver 50% to 100%
> > > > more performance per watt and per dollar, particularly
> > > > when considering the additional chips that Intel needs to complete the server design.
> > > >
> > > > Therein they are mentioning the advantage of SoC vs a CPU. As I said before, the 80W are for
> > > > the whole SoC. The 95W are only for Xeon CPU, adds the TDP of rest of components to the
> > > > Intel platform and you will need up to double power to do the same work than ARM SoC.
> > >
> > > Your calculations are wrong, but your point is good.
> > >
> > > Intel does not (currently) integrate networking, and that leaves an opportunity for a competitor to offer a
> > > different system architecture that is differentiated. That
> > > being said, it only costs about 10-20W to add 4x10G
> > > ethernet MACs. I'm not sure about the actual $ cost, but Intel would simply need to lower their prices.
> > >
> >
> > Add the south Bridge and now consider the performance gap, 350 vs 320, and you will obtain that you
> > will need something close to twice more power to do the same work than the ARM SoC. I got 89%.
>
> Let's compare calculations here.
>
> I've got:
>
> ThunderX: 80W @ 350 SPECint_rate
> Xeon E5-2650L: 159W @ 546 SPECint_rate, including 2x70W CPUs + IOH + 7W for a dual
> 40G ethernet controller, note that you can fit upto 5 NICs for each server CPU.
>
> However, Thunder will probably be compared against HSW-EP or BDW-EP,
> which should be about 30-50% better performance/watt than IVB-EP.
>
> David
We were discussing the chips in the table 1 of the Microprocessor Report article about ThunderX. My above 89% is for the model of Xeon listed in that table. Now you change the model...
I will assume that your computations for the new Xeon model are right. Even assuming that, the TunderX is about 30% more efficient.
Yes, ThunderX will be competing against Broadwell-EP. Taking your best value of 50% improvement. This means that an expensive Broadwell-EP Xeon on 14nm FinFET will be only slightly better in performance per watt than a cheap ARM SoC made on 28nm planar.
But I already said you that ThunderX is not where the real battle will be. The real battle will involve the X-Gene v3, Vulcan, and K12, all them will be made on 14/16nm FINFET.