anon (anon.delete@this.anon.com) on November 27, 2013 3:32 am wrote:
> Michael S (already5chosen.delete@this.yahoo.com) on November 27, 2013 2:25 am wrote:
> > anon (anon.delete@this.anon.com) on November 26, 2013 6:42 pm wrote:
> >
> > > That comes out at 1.05W per core at 2.4GHz (presumably). Increase frequency to
> > > 2.6 probably brings power up significantly, but it seems 72 cores could fit into 100W.
> > >
> >
> > Which still too high, considering 200W TDP for the whole chip.
>
> How much would wider vector units add, in your estimate?

I have no idea, but my estimate would be that for KNC the ratio of the power consumption of VPU and of associated wide data pathes to/from L1D to the rest of the core is 4:1 or 5:1.

>
> How would the power breakdown fit if there are 1.3GHz clocked cores?
>

14nm Silvermont-derived core without VPU at 1.3GHz? I'd guess around 0.4W. But my estimate based on little more than nothing :(