The way you word that suggests that overclocks higher than 2200mhz are unlikely to happen because of the voltage lock, I don't think that is the case. Overclocks over 2200mhz are probably going to be rare, and like Maxwell will favor lower voltages. The "thing" that makes ICs fail at high clocks is closely tied to temperature; there are also architectural factors that affect the clocking. For the sake of the argument let's imagine the GPU as a huge 2D matrix that represent gates. The GPU is a hugely parallel machine, so you can imagine that the thousands of independent paths (from input (top) to output (bottom)) represent the longest datapaths possible. The longest delay between any two "latched events" (when data must be "pulled" from a rail at a clock edge trigger) must naturally be shorter than the inverse of the clock rate by a reasonable amount. If not there will be confusion, errors, crashing and lots of frustration. Other than the physical propagation time of the signals (hence path optimization in pascal to allow for 2ghz+) you also have to factor that transistor gates themselves present an inherent latency called "gate delay" that increase with temperature; but I'm not taking about the GPU temperature. Local temperatures; "instantaneous hotspots" will affect gate delay before your average cooling solution can dissipate the heat away. This is why LN2 will probably be the only way forward, because much lower average thermal energy (overall chip temp) means local hotspots will dissipate more quickly What I have written above is why even on water 1600mhz 980Ti is hard because you want to keep it below 50c ideally, of course it depends on the chip and the voltage you need to sustain such a clock, but as a general rule if you can keep the GPU core and VRMs cool you will have more success for those overclocking 'fringe benefits' At the end of the day when someone complains that GTX 1080 "sucks for overclocking" you should realize it's a very petty complaint, because it's entirely contingent on where on the clock scale NV decided to stick the "stock" mark. The stock boost clock is 1733mhz. They could simply lock the clock at 1733 and not let it rise above that. You could overclock it to 2100mhz and everyone would be impressed. Instead it clocks out the box to 1860mhz so you're less impressed with the OC but the performance is the same.