Article: Power Delivery in a Modern Processor
By: Yoav (no_drink.delete@this.gmail.com), May 13, 2020 3:56 am
Room: Moderated Discussions
If not all domains under the save Voltage regulator need the same voltage, LDOs can save a lot of power.
Example:
Domain1 need 0.8V, and Domain2 need 1.2V.
So VR gives 1.2V.
Without LDO, domain1 would get 1.2V, and burn a lot of power.
With LDO, it can use 0.8V, and burn less power.
David Kanter (dkanter.delete@this.realworldtech.com) on May 12, 2020 10:16 am wrote:
> Travis Downs (travis.downs.delete@this.gmail.com) on May 12, 2020 7:53 am wrote:
> > A nice read!
> >
> > What is the purpose of the LDO? Does it even reduce power draw? The voltage drop is a dead loss
> > in an LDO, in the sense that the reduced power I * (Vin - Vout) is just dissipated in the regulator,
> > but is there a secondary effect that saves power? Or is it just to better control the voltage
> > for another purpose (one could be to move power dissipation away from the core)?
> >
> >
> >
> >
> > I believe "power" should be "voltage" in that sentence.
>
> Thanks for the correction. I fixed that.
>
> LDO's save a bit of power. Since switching power is ~V^2, and the power dissipated by an LDO is linear WRT
> voltage drop, you can improve power with an LDO. So your power savings are equal to the resistive losses.
>
> In contrast, a buck converter has a roughly constant efficiency. So the regime where an LDO
> helps is small swings (say
> David
Example:
Domain1 need 0.8V, and Domain2 need 1.2V.
So VR gives 1.2V.
Without LDO, domain1 would get 1.2V, and burn a lot of power.
With LDO, it can use 0.8V, and burn less power.
David Kanter (dkanter.delete@this.realworldtech.com) on May 12, 2020 10:16 am wrote:
> Travis Downs (travis.downs.delete@this.gmail.com) on May 12, 2020 7:53 am wrote:
> > A nice read!
> >
> > What is the purpose of the LDO? Does it even reduce power draw? The voltage drop is a dead loss
> > in an LDO, in the sense that the reduced power I * (Vin - Vout) is just dissipated in the regulator,
> > but is there a secondary effect that saves power? Or is it just to better control the voltage
> > for another purpose (one could be to move power dissipation away from the core)?
> >
> >
> >
One advantage of the FIVR is that the power from the motherboard VRMs
> > to the processor is about twice as high as in a conventional system.
> >
> > I believe "power" should be "voltage" in that sentence.
>
> Thanks for the correction. I fixed that.
>
> LDO's save a bit of power. Since switching power is ~V^2, and the power dissipated by an LDO is linear WRT
> voltage drop, you can improve power with an LDO. So your power savings are equal to the resistive losses.
>
> In contrast, a buck converter has a roughly constant efficiency. So the regime where an LDO
> helps is small swings (say
> David