CPU voltage regulator integration

[Jamie M]

Hi,

Apparently some new Intel CPU's are being tested with on board voltage
regulators (2.4V input buck converters for CPU power rails). Whats
really surprising is these voltage regulators include the power mosfets
and also the inductors, and possibly even the capacitors I'm not sure!

The efficiency right now is only about 80% for the on die regulator,
which might be a deal killer I'd say, plus it is eating up expensive
real estate on the CPU die, and adding heat. But I wonder how long
until IR.com has a programmable one of these on their site?

Here are the links to the story and a pdf of the technology:

http://hardware.slashdot.org/story/13/05/13/2249209/intels-haswell-moves-voltage-regulator-on-die

http://www.psma.com/sites/default/f...ully-integrated-silicon-voltage-regulator.pdf

cheers,
Jamie

 

[Jasen Betts]

Hi,

Apparently some new Intel CPU's are being tested with on board voltage
regulators (2.4V input buck converters for CPU power rails). Whats
really surprising is these voltage regulators include the power mosfets
and also the inductors, and possibly even the capacitors I'm not sure!

The efficiency right now is only about 80% for the on die regulator,
which might be a deal killer I'd say, plus it is eating up expensive
real estate on the CPU die, and adding heat. But I wonder how long
until IR.com has a programmable one of these on their site?

80% is not a big deal, the current crop of intel CPUs have tiny
heatsinks, and low heastsink clamping force compared to 4 years
ago. They can't be producing much heat to start with. Another
20% isn't going to be a big deal.

 

[rickman]

Hi,

Apparently some new Intel CPU's are being tested with on board voltage
regulators (2.4V input buck converters for CPU power rails). Whats
really surprising is these voltage regulators include the power mosfets
and also the inductors, and possibly even the capacitors I'm not sure!

The efficiency right now is only about 80% for the on die regulator,
which might be a deal killer I'd say, plus it is eating up expensive
real estate on the CPU die, and adding heat. But I wonder how long until
IR.com has a programmable one of these on their site?

Here are the links to the story and a pdf of the technology:

http://hardware.slashdot.org/story/13/05/13/2249209/intels-haswell-moves-voltage-regulator-on-die


http://www.psma.com/sites/default/f...ully-integrated-silicon-voltage-regulator.pdf

That sounds a lot like the Enpirion technology. They have been selling
switchers for a number of years that include on die magnetics. I agree
that it seems an odd duck to want to include in the CPU package much
less on the die.

 

[[email protected]]

It looks like the switcher is a separate piece of silicon, in the packagewith
the CPU but not the same chip.

maybe that is because it is a test chip?, it says 90nm, the processor
they used is 65nm. don't know if it is even possible to do all the
thick metal and magnetic layers in the process used for the CPU

the e7330 they used is already two processor dies in one package

The umpta-phase switcher is cool. I've done three phase, but IC designersdon't
much care about parts count.

What's the "virus" on p18 of the PDF?

I believe that is what they call a special software that can put
maximum load on all parts of the chip at the same time

-Lasse

 

[josephkk]

Hi,

Apparently some new Intel CPU's are being tested with on board voltage
regulators (2.4V input buck converters for CPU power rails). Whats
really surprising is these voltage regulators include the power mosfets
and also the inductors, and possibly even the capacitors I'm not sure!

The efficiency right now is only about 80% for the on die regulator,
which might be a deal killer I'd say, plus it is eating up expensive
real estate on the CPU die, and adding heat. But I wonder how long
until IR.com has a programmable one of these on their site?

Here are the links to the story and a pdf of the technology:

http://hardware.slashdot.org/story/13/05/13/2249209/intels-haswell-moves-voltage-regulator-on-die

http://www.psma.com/sites/default/f...ully-integrated-silicon-voltage-regulator.pdf

cheers,
Jamie

Cute. I expect that the shebang is more in package rather than completely
on chip. More likely the regulator is on chip but the switch, inductor,
and caps are in package. I'll bet it runs at some several MHz.

?-)

 

[[email protected]]

Cute.  I expect that the shebang is more in package rather than completely
on chip.  More likely the regulator is on chip but the switch, inductor,
and caps are in package.  I'll bet it runs at some several MHz.

?-)

it's right there in the presentation, it is a separate chip but
switches and inductors are on chip and it is programmable between
30 and 140MHz

-Lasse