So I had my phone OC'd to 780 and it was pretty zippy and was loving it but after 8.5 hours of moderate use, I was down to around 30%. Today I dropped to 710 after a post stating that someone killed their CPU by using a setting that high and I just noticed that after 8.5 hours, I currently have 62% of my battery left..
just saying... big difference for little gain
Your observation is absolutely correct... here is the explanation.
In a simple, fixed voltage processor which dissipates no power when the clock is not running, the power dissipation is proportional to the clock frequency. We might say something like:
P ~= k*f
Where 'k' is a constant and 'f' is the clock frequency. In this scenario, if you reduce the clock down to zero, the power dissipation goes to zero as well. Also, the total energy used by a program that always does exactly the same sequence of instructions every time it runs is fixed - no matter how slowly or rapidly the clock is ticking, running that program will use up a fixed amount of energy from the battery. The only choice you have in the matter is "do I want the result sooner or later"?
If you dig down a little further, you might ask, "but what is this 'k' factor?". In doing that, you would find for the simple, fixed voltage processor (with no clock gating or other fancy power management tricks) that
k = C * 0.5 * N * V^2
where N is the total number of logic gates, C is a constant relating to the manufacturing process, and V is the logic supply voltage.
So the Power dissipation equation is now
P ~= C * 0.5 * N * V^2 * f
It's that Voltage-squared term (V^2) which is causing the results you are observing. The reason? Our Eris uses successively higher logic supply voltages as the clock frequency is stepped up. This is done because detection of logic levels is always done in the presence of noise, and as you squeeze the logic timing tighter and tighter with increasing clock frequencies, you need a larger logic voltage swing to be able for the logic gates to run correctly - especially at the moment in time when you change the clock frequency.
This is unfortunate - because it means that in reality, running faster means that our hypothetical computing job
costs you more in Energy (battery reserve) when running quickly than if you ran it slowly. In very real terms, you are "going faster, but getting less work done" per mAh of battery capacity.
In reality, what goes on with the processor relating to power dissipation is far more complicated than what I've indicated here, but that is a very rough, first-order approximation.
To see how this works, imagine you are running at a given frequency, and suppose that the next step up in frequency is 10% higher, and that also the supply voltage is increased by 10%. Well - you will get 10% more computational work done per unit time; but the power will increase by
(1.10)^2 * 1.10 = 1.33 ... 33%!
Speed kills (batteries)!
eu1
(oh, yeah, you know who you are).
