By: David Kanter (dkanter.delete@this.realworldtech.com), August 17, 2014 11:31 am
Room: Moderated Discussions
Ricardo B (ricardo.b.delete@this.xxxxx.xx) on August 17, 2014 11:37 am wrote:
> Michael S (already5chosen.delete@this.yahoo.com) on August 17, 2014 11:04 am wrote:
> > David Kanter (dkanter.delete@this.realworldtech.com) on August 17, 2014 10:23 am wrote:
> > >
> > > 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.
> >
> > MACs or MAC+PHys?
> > I have no experience with 10GE MACs, but if they are even remotely resembling 1GE MACs then 2.5-5
> > W per MAC at 22 nm sounds way way too high even including XAUI interfaces to off-chip PHYs.
>
> Yep, that power level is more compatible with a MAC plus a 10GBASE-T PHY.
>
> > As to 10GbE PHYs, is it technically possible to integngrate them on the same die with high-performance CPU?
> > Again, without knowing much I'd guess that integrating 10GBASE-KX4
> > is possible, any other variant of 10GbE will be problematic.
>
>
> Integration of the SFP+ interface is probably feasible.
> SFP+ direct attachment remains a cost effective alternative to
> 10GBASE-T in the data center and is prefered to 10GBASE-CX4.
> Integration of 10GBASE-CX4 is also possible, but not interesting.
>
> Hmm.. maybe we should just stick with PCIe and let the client choose the NIC?
>
So Intel has an ethernet controller that can do 2x40g or 4x10g ports. It's a 7W part, but I have no idea whether that includes the PHY (or which PHY).
David
> Michael S (already5chosen.delete@this.yahoo.com) on August 17, 2014 11:04 am wrote:
> > David Kanter (dkanter.delete@this.realworldtech.com) on August 17, 2014 10:23 am wrote:
> > >
> > > 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.
> >
> > MACs or MAC+PHys?
> > I have no experience with 10GE MACs, but if they are even remotely resembling 1GE MACs then 2.5-5
> > W per MAC at 22 nm sounds way way too high even including XAUI interfaces to off-chip PHYs.
>
> Yep, that power level is more compatible with a MAC plus a 10GBASE-T PHY.
>
> > As to 10GbE PHYs, is it technically possible to integngrate them on the same die with high-performance CPU?
> > Again, without knowing much I'd guess that integrating 10GBASE-KX4
> > is possible, any other variant of 10GbE will be problematic.
>
>
> Integration of the SFP+ interface is probably feasible.
> SFP+ direct attachment remains a cost effective alternative to
> 10GBASE-T in the data center and is prefered to 10GBASE-CX4.
> Integration of 10GBASE-CX4 is also possible, but not interesting.
>
> Hmm.. maybe we should just stick with PCIe and let the client choose the NIC?
>
So Intel has an ethernet controller that can do 2x40g or 4x10g ports. It's a 7W part, but I have no idea whether that includes the PHY (or which PHY).
David