BLAfH

Direkt nach dem Studium in die Schweiz ist ziemlich schmerzfrei, das habe ich vor gut 10 Jahren gemacht. Informatiker sind gefragt, es wird auch aktuell noch rekrutiert was sich findet, auch aus dem Ausland. Und EFTA-Raum ist in Sachen Bewilligungen - sei es Bewilligung B/C zum wohnen in der Schweiz, oder Grenzgänger - auch kein grosses Problem. Lebenshaltungskosten vs. Lohn vs. Steuer und so weiter muss man halt mit spitzem Bleistift durchrechnen.

Because developers for those consoles simply don't use those new instructions.
 
Those instructions are optimizations to solve certain problems in a faster way. They won't allow you to do something completely new. So by not using them, you loose some performance. Which isn't that much of a deal in console development, as they mostly target 30 to 60fps anyway, and can optimize more than on PC (as they know exactly what hardware is inside a PS4, not like a PC where you have to support AMD and Intel CPUs, DualCores to 32 Cores, Radeon and GeFore, and so on..).
 

Simplified to a point a hope is juuuust before it gets "wrong", but given the point to start...: 
 
All of them matter. Every one of them can kill something, but the main suspect here is IMHO overall wattage.
A VRM is, as it's name implies, a module, not a single part. There's controller chips, MOSFETs, coils, capacitors, the lot.
Each of them can be killed by one thing or the other, all of them will be killed by heat which ultimately burns down (pun intended) to watts. 
It's next to impossible to kill a VRM with it's output voltage, as you can't set it to one that harms it. You may set it to some thing that kills a CPU, but you can't set it to a voltage that harms the VRM in and itself - regarding only the voltage.
 
Amps - well, there is an upper limit of amp, some of the component can handle. Although this is - in most cases like MOSFETs - not only amps, but in a big part amps at specified voltage -> therefore Watt. 
 
Some very rare effects ignored, it boils down to efficiency and temperature. VRMs take a "high" voltage (12V) and convert it down to a lower, in your case... 1.3 to 1.4V for the CPU. It can't do that with 100% efficiency (well, physics, and unfortunately there are no room temperature superconductors in sight) so there's always something lost. And any loss in a electrical system is... heat. 
 
Let's assume your VRM is - over all - 90% efficient (simplified view on the VRM as a whole, no regard for varying efficiency at different  loadpoints and voltages and on different parts of the vrm, including the disadvantage of low volt high amp regarding On-State Resistance in FETs, cross heating of components  etc.pp.,  wildly guessed number, just for a nice calculation, results rounded). Cooling on the VRM is good to keep it in spec for max 15W heat generated at the VRMs.
 
Your CPU is pulling 1V@100A. 100W. 10% is "lost". So your pulling 111W on the PSU side. 11W will be converted to heat. All is nice and dandy.
Another CPU is pulling 2V@50A (you won't find one in real life, but... for the calcs, meh...). As above... 100W, 111W on the PSU side, 11W in heat. all is nice and... 
 
You see? Neither 100A nor 2V alone have killed anything.
Well... Put in a CPU 2V@100A. 200W. 222 on the PSU side. 22W in heat. You can cool off only 15.
That's about 50% thermal overload, you fry some part on your VRM. Or, more likely... you degrade them. 
Capacitors for example have a lifetime of... lets say 5000h @ 85C max. - overheat them... and they may only have 50.. at 170C. 
 
It won't be "bang, dead". It will be more like... runs a day... Bluescreen. runs 4 hours... Bluescreen. runs  15min,... dead 
 
And it also shows how you can improve the design of the VRM. You can have more "phases", sharing the load and heat over more components, of which each one won't get as hot. Cons? Cost and space. 6 Phase cost 50% more than 4. And it needs room on the board, and it has to be in a certain spot on the board, limiting space and cooling overall again...
You can improve cooling (to a point... e.g. the MOSFETs in your VRM are tiny pieces of silicon, and there's a limit on how much heat you can remove from such a small area). Again. Cost. Space. 
You could make a more efficient VRM - but those VRMs nowadays are already very close to whats doable. 
 
Also as a side note on cooling VRMs: If you're thinking about a AIO for the CPU... Those tend to be worse for VRM temps, as the don't generate airflow on the VRM area on the board as a air-cooler does. Cooler CPU, higher OC. Dead VRMs. 
  
 

What? I get the impression you're massively misunderstand overclocking. 
 
You are guaranteed stock speeds. And not a singe MHz more. 
Anything else is luck. And your own responsibility, should you destroy your CPU or mainboard.
 
AMD is just friendly enough not to forbid you to take this responsibility (as Intel does on most CPUs by locking them to stock) or ask you to pay more to do it (as Intel does by selling unlocked "K" chips for a bit of a premium.) 
 
CPU and mainboards are tested for stock speeds (+ ab bit of headroom for security margin).
Unlocked means, you are allowed to try to get into that margin and use it.
 
No luck? We'll... pitty. But nothing you can complain about. 

There is no "best". There may be a "ok for this job". 
If that's BSD, Linux (whatever disto fits the bill), Mach, Windows, RTOS or.... whatever -  depends on "the job".

I've usually use the add on bundle "Users". Has a random function that comes up with all kinds of errors at the most unfortunate times.
About every other month I'll get the premium bundle "external engineer", which can have power over 9000 (in people not able to work after he changed "nothing").