Originally Posted by hak8or
Ok, there is no way that they can fit the entire Voltage regulator onto the chip. The passives, such as inductors and capacitors, stay on the motherboard because of their size. You cannot make smaller inductors and capacitors without increasing the switching frequency drastically, which will in turn generate lots of heat and eventually become very inaffeciant. The difference haswell will make for the VRM, or voltage regulators, is that they put the controller, the thing that says "ok, the voltage is 20 millivolts too high, fix that now" onto the chip.
Here is an example of a voltage regulator used specifically for processors. http://www.irf.com/product-info/ipowir/fs8063.html
And an example of a full blown buck converter (SMPS). http://www.irf.com/technical-info/refdesigns/irdcip2003a-c.pdf
Why this is better than having discreet voltage regulators I do not know, all reason would tell me that it would actually worsen the quality of power significantly because a very good voltage regulator takes a significant amount of space on the chip, something Intel does not want. Not to mention we do not know how many phases the voltage regulator supplied by haswell will have. Will it be 4, or 12, or allow for more phases via a dedicated chip? Yes, power efficiency will be very good, but I actually see this as very threatening towards overclocking, because the quality of power to the chip may be decreased. It does make it cheaper for motherboard manufactures though, since they do not have to include or worry about a voltage regulator.
If anyone is interested, the voltage regulation scheme used for processors is called a buck converter. http://en.wikipedia.org/wiki/Buck_converterIf you want to supply a lot of current, then you use synchronous buck converters, otherwise known as a buck converter with multiple phases.
Multiple phases mean that you have a pair of mosfets (switches) and an inductor per each phase.The more the phases, the better response time (normally) and less strain on each mosfet and inductor. It splits the demand per phase so to speak. That is why when you look on a motherboard near the processor you often see little cubes, those are the inductors. If the motherboard is designed for overclocking or hefty processors, both of which require large currents, then the mosfets have heatsinks on them.
I am an aspiring EE, so feel free to ask any more questions, I would be happy to answer them or link to more information!