By: Paul A. Clayton (paaronclayton.delete@this.gmail.com), September 20, 2012 2:53 pm
Room: Moderated Discussions
somebody (delete.sombody.delete@this.somebody.com) on September 19, 2012 9:27 am wrote:
[snip lots of interesting information]
> With respect to your idea to use capacitors spatially and temporally
> to handle peaks, the parasitic resistance of your lateral power
> distribution comes into your way: Decoupling is of no good use if not
> close enough to where you need it => Dynamic spatial allocation does
> not solve the problem. If the decoupling is close, it's there anyway
> occupying the area => The dynamic temporal allocation is void.
Thank you for the explanation. I was merely thinking that there might
be a way to allow a small excess of power to accumulate into an energy
store which energy could then be used in uncommon events to meet timing requirements without (or with less frequent) replays. (ISTR that at least one paper had indicated that the power efficiency of an ALU could be improved if the worst case timing requirements could be relaxed, even using replay to handle the exceptional cases could result in a net increase in energy efficiency. If the need for an extra boost of power to increase speed and meet timing could be known--or accurately predicted--early enough, replay could be avoided. Since the power storage and delivery alone is impractical, this thought can be tossed in the circular file.)
Thank you also for the other information and the link.
[snip lots of interesting information]
> With respect to your idea to use capacitors spatially and temporally
> to handle peaks, the parasitic resistance of your lateral power
> distribution comes into your way: Decoupling is of no good use if not
> close enough to where you need it => Dynamic spatial allocation does
> not solve the problem. If the decoupling is close, it's there anyway
> occupying the area => The dynamic temporal allocation is void.
Thank you for the explanation. I was merely thinking that there might
be a way to allow a small excess of power to accumulate into an energy
store which energy could then be used in uncommon events to meet timing requirements without (or with less frequent) replays. (ISTR that at least one paper had indicated that the power efficiency of an ALU could be improved if the worst case timing requirements could be relaxed, even using replay to handle the exceptional cases could result in a net increase in energy efficiency. If the need for an extra boost of power to increase speed and meet timing could be known--or accurately predicted--early enough, replay could be avoided. Since the power storage and delivery alone is impractical, this thought can be tossed in the circular file.)
Thank you also for the other information and the link.
