New CPUs with Vreg on the die

[Psi]

http://hothardware.com/News/Haswell-Takes-A-Major-Step-Forward-Integrates-Voltage-Regulator/

Assuming its user adjustable (or hackable) it should make overclocking better since one of the limiting factors currently is vdrop under load causing the cpu to crash.

Normally you have to be careful that the voltage isn't too high at 'idle' and doing so limits how high you can go under full load voltage.

[mariush]

Here's the pdf with the slideshow intel presented : http://savedonthe.net/download/486/400a-fully-integrated-silicon-voltage-regulator.html

originally here but it might be slashdotted : http://www.psma.com/sites/default/files/uploads/tech-forums-nanotechnology/resources/400a-fully-integrated-silicon-voltage-regulator.pdf

Basically, the motherboard will only do 12v to 2.4v and the vrm in the cpu will do the rest.

[peter.mitchell]

i'm unsure how to feel about this.

[Psi]

I wonder why they picked 2.4V
3.3V would have made more sense. It's already provided by the psu and wouldnt need MB regulators at all.

Perhaps there are issues drawing high currents on such a low voltage psu rail.

[Razor512]

I wonder why they picked 2.4V
3.3V would have made more sense. It's already provided by the psu and wouldnt need MB regulators at all.

Perhaps there are issues drawing high currents on such a low voltage psu rail.

Many power supplies only offer around 70-80 watts on the 3.3v rail while a high end CPU can easily pull 120+ watts, and when overclocked, can pull 200+ watts, so since the 12V  rail is the most powerful one offered by the power supply, it is often used to power the CPU.



-==-=-=--=-=-=-=-=-==--=
hopefully someone can do some stress testing using air or liquid cooling and not liquid nitrogen like all of the current overclocks found are doing.

VRM's are the most annoying part to deal with when overclocking, if the motherboard has crappy ones, you get this problem

MSI 870a-g54 VS a 125 watt load

Hopefully the CPU can handle a decent amount of voltage and a good overclock and not fail after extended heavy use.

[Hypernova]

Pretty freaky power density, not to mention the insane frequency they are running the switchers at.

But how are they going to deal with the extra converion losses the tiny CPU package will now have to carry?

[amyk]

Great, another heat-producing part gets put under the heatsink... although maybe this will reduce heating at the CPU pins since they can carry less current.

Modern CPUs are amazingly high current devices; 100A+ at around 1v.

[Fraser]

It's a negative vote from me on this. If there is one thing I have learnt over the years it is that power supplies and regulators are the cause of many a failure in electronic systems. The last thing I would want is a power regulator on the CPU as a high risk failure point. I would rather have all the regulation on the PCB under the CPU, or close to it, but not inside it. Remember the old Pentium II....it had all the clever stuff crammed onto a PCB that formed the 'CPU'.

[tom66]

So, soldered on CPUs, and now the CPU integrates the VRM. Modern motherboards won't be repairable any more. Sad day.

Also, a good way to stop people over-volting the CPU, and gaining a bit more performance.