Now then. While X86 desktops differ in application and language, than say a PowerPC, or RISC based platform, I'd like to say that this thread will make broad generalizations that I like to think apply to both.
I'll try not to make too many assumptions; so I'll get those out of the way to begin with. I'll be assuning that those whom are reading this and overclocking phones, have not done so with desktops. If your strictly a handset techie, then this is perfect for you.
I'm trying to apply my experience and endeavors to another practice with cormidbalities.
I inherently feel I need to clarify my intentions, as back on my home land, on OCN, we're a very liberal, and very aggressive, flaming-prone board.
I've taken an i5 661 to 6.86ghz. A former world record holder. -191c during a DICE run. Even was bold enough to do a twin-stage phase unit run @ -2c.
Taken more Nehalem/Lynfield, Gulftowns and Sandybridges to the upper limits of what some say simply aren't possible.
Unlike the above body of text that either strokes my e-peen, or no one gives a (insert your favorite f word here) about, I'll make the actual mean of this thread a bit.... Short.
Heat kills faster then voltage. Electronics inevitably die from one of two invariable conclusions. Electromigration, or heat death.
I won't delve into thermodynamics, but simply put, as the electrons in the SoC's gates heats up (die heat) the electrons move faster, thus accelerate electro migration. (They're one in the same. To a degree)
I should clarify what electromigration is. Put simply its where electrons eat away at the pathways for the flow of an electrical current. Thus mitigating capacistance for transient flow of a charge. (Akin to errosion)
High amounts of heat (defined as above the TDP, or thermal design power) accelerate this ocourance.
Or high amount of heat in general can fuel it. Or it can cause warping of gates, disallowing flow. And a plethora of other issues.
What I'm saying is that, assuming you keep the SoC within the TDP (yes. You may overvolt; assuming you're within voltage threshold (check the manufacture for details) you may "safely" go above this, assuming heat is kept in order.
Keep either heat low, and voltage, high, or vice versa, for a high OC. (Over clocking is never truly safe. Think rooting your $600 phone is hard? Try pushing your 980X to 5.0ghz on air and not killing it. (They were $1,000....)
I'll update this later as there's more. Last thing for now is, power.
A clean, and stable supply of power is key. You do not want voltage, wattage, amperage, fluctuating. You want to keep a specified output rock solid.
Typically a solid state capacitor, and not a dialectic based is preferd... But I digress. This is too akin to desktop.
As for a bit of foreground, we overclockers use sub-zero cooling methods, because we slow the flow of the electrons so much, so, that we get out CPU to last, what normally would be a few seconds, (to as soon as I push the power button; with a combination of dry ice, liquid helium, and liquid nitrogen, and getting as low as -198c, we can get the chip to live for about 3 hours... Typically a good benching session.
Electromigration becomes evident at about 2 hours. (I'll know because it'll refuse to boot, and I'll have to increase voltage... Seems sane when I'm already at 2.0V with no high-K gate (high K gates did a dream for the overclocking community. It made chips handle high heat like no other...) But made them more sensitive to voltage.
Sent from my SM-N900P using XDA Premium 4 mobile app