Microsoft does not give you EXPLICIT instructions, or any instructions to format the system drive on a running operating system.
Here’s why this is statement above is problematic, though I fully understand what you are trying to achieve with this (In general I will explain this for other people too, not just yourself.)
Summary
If you want to get into the specifics of designing a digital power supply for an MCU or CPU (any type of general purpose processor,) you have to start with the technical reference manual to get the actual supply characteristics (whether it be single or dual supply.) You cannot specify just generic ranges for ALL processors. Most of these also have on-chip voltage regulators, internal frequency reference, etc, so the ranges are determined entirely based on the input characteristics (power integrity of the entire signal chain, it’s not just the ATX power supply, every motherboard has it’s own digital SMPS / switch-mode power supply.) You can’t draw too many conclusions about one voltage range, given it’s dynamically scaled based on the processor load. To actually throttle or undervolt / overvolt, in a generic way for all processors would not work (given you have to use values from Intel’s technical reference manuals, which can be found on the Intel ARK site / app.) One thing you want to avoid is trying to undervolt or overvolt, given the processor largely manages / schedules the workload by itself (it will throttle itself based on the workload, and this is also an aspect of DYNAMIC power management.)
If you throttle it based on a voltage setting, because you think that’s a problem, it’s going to have to work really hard to get the same thing done, which in turn will just make it max out all the time, and also waste power. It has to be able to finish the workload on an acceptable timeframe to be able to switch power modes effectively (so the workload is associated with the processor type itself.) In some situations you would actually just be creating underlying issues, which actually work against your goals. There are a number of variables that determine the supply characteristics (I would check NOT ONLY the datasheets but ALSO the technical reference manuals.) The stability of the input source, depends more on the motherboard, than it does the processor itself. You would need pages upon pages of text to get into this, it’s very specific to the OEM that produces the motherboard, not so much Intel. In general, don’t do this, only use the official errata / documentation on this processor (from Intel,) which is publicly available. If you throttle it heavily, it will increase the processor workload, it will take longer to get the same job done, and it will be inefficient, just because it’s unable to scale the frequency dynamically and allocate resources accordingly (then scale back down into an intermediate or low power state.) Voltage stability and temperature stability in some ways have a lot of parallels, to change the input supply characteristics, aside from what is set by the OEM is a very dangerous. I mean you can go into every single factor which relates to trace impedance, process node, the overall die size, even just the power architecture itself and supply characteristics, one set voltage does not inform you that there is a problem by itself (it’s a digital system.)
I mean to think about this more, no two devices will be exactly the same, some might even be rebadged versions of other SKUs (which have features or cores disabled, so the supply characteristics might not tell the whole story.) You would have to have a simulator to achieve the results you want (IBIS / full wave EMF simulator.) That could take you 2 weeks easily with several 8-core processors and massive memory banks, working 24/7 (I mean for most 8-16 layer multi-layer PCBs like this. It has a lot of transceivers too just for the PCIe network, which is of course packet-switched, and in most cases uses the same encoding scheme on the PHY, as the USB / SATA channels.)
I would look at testing reactive components on the board first, like capacitors, to see if they have failed (If you’re worried about that, given they have lower temperature stability than the other components.) If you really seriously want to explore this topic, you would have to look at the technical reference manual to find acceptable voltage ranges / values for this SKU, but also the fully supply requirements (in most cases 1/3rd, or even half of the lands on the PCB could be devoted SOLELY to supply requirements. You would need a PMIC with an array of multi-channel supply sequencers, just to bootstrap the processor. For processors with a large interposer, high ball count, easily a few hundred lands could be devoted JUST to the power supply. It’s hard to say, because this is OEM-specific (and they all choose different layouts, have different target markets, most being embedded devices, maybe 2% being applications processors like this.) Changing a static input value, such as voltage, on any one of the rails, doesn’t translate directly into results, and will most likely hamper any of your efforts.) Trying to debug this based on a single reading would be ridiculous (there would be visual indicators that would be present in most cases on older motherboards. Newer ones utilize what are known as “Conductive Polymer Hybrid Aluminum Electrolytic Capacitors,” for the digital power supply, which have far greater temperature stability and in turn the lifespan of the component is greatly extended, due to the dielectric material and housing being comprised of different materials more suited to the operating environment. You would need a good ESR or LCR meter to test these, and it’s not worth getting into unless there are massive variances which would interfere with the operation of the digital power supply itself. In some cases it could result in flipped bits maybe, but it would be able to recognize this on its own.) There are too many problems with this approach, at least for it to be reliable in the short term. The datasheets and accompanying technical reference material (processor errata in general,) will list all the acceptable tolerances and ranges, based on revision / stepping / model / etc. There is no one set perfect range for everything, and that’s probably the short answer (The answer is to read a lot, make sure your decisions are informed based on the reference material, make informed decisions based on actionable information contained within the reference material.) Also recognize that the components drift in value with age, it doesn’t always matter to look at it purely based on a static value in a datasheet.
Real world values depend entirely on characterizing the device itself (there is some margin of error in these numbers stated in the reference material / datasheets.) You would have to factor in all aspects of this before proceeding or taking any steps. This is entirely outside the realm of IT work (more getting into electrical engineering, electronics technology, etc.) You couldn’t do this with some amount of intermediate design data from the OEM, some of which is proprietary, so of course you would have to go off of IBIS models (which legally you are allowed to do that, but you will still come up short because you don’t have a netlist with trace impedance values, various dimensions of the copper pours, any type of via stitching or characteristics relating to the dimensions and layouts of lands buried beneath certain layers which are not visible to the naked eye. The entire PCB stack itself follows a pattern also, could be any variation on power, ground, and signal planes, you don’t know which areas have analog and digital grounds, if it’s hidden somewhere, or what features of the processor are left unused, whether it’s routed to a ground and terminated with a bypass cap / pull up / pull down resistor, in what order you are specified to do this based on the reference documentation. You don’t know what features the OEM has enabled on the motherboard for this specific iteration.) So you’re basically done, it won’t work without a monumental amount of resources and effort, if you are after quick results that is. It’s way smarter to just look for more obvious indicators on the board itself, or try to find another indirect way of influencing the behavior of the processor (like changing the maximum frequency in the power profile on the operating system itself.)