I did also say that "Power use for different speeds is not necessarily linear." (I underlined it for sake of clarity this time.)
Originally Posted by clevin
If there is an article there specific to uberkernel, I can't find it using the following search string in Google: site:webos-internals.org uberkernel battery -inurl:git
In any case, battery life in a cell phone is an extremely subjective thing and subject to many factors including radio availability, interference, reflections, and refractions, not to mention all the varying software use that may be going on. Until someone undertakes some dedicated time testing, all of the evidence is anecdotal.
You might want to check your facts since modern CPUs do vary their voltages based on the speed of the moment. That goes back to the late 1990s with early SpeedStep and PowerNow! implementations in mobile chips and has been fairly standard for a number of years now in desktop chips as well. While a given CPU may clock higher with factory voltages (higher even than the factory-labeled max speed of the chip), that doesn't mean that there are no voltage changes going on.
Higher voltages is NOT required for OC up to certain clock.
The reason for this is that as the chip gets hotter, its electrical resistance rises, hence requiring higher voltage to get past the resistance. This, in turn, creates more heat.
I even found this article from 1997 that discusses manual voltage changes for overclocking the Pentium and Pentium Pro chips, which is around when I got my start. The technology has only spread since then, as we've been seeing in modern mobile chips.