a question about undervolting

[robinta]

I have undervolted on my S2 for a while now, using different kernels and have a question i cant see answered anywhere else.

Does undervolting (i.e applying a reduction to each frequency) or reducing the settings for the gpu result in the best battery savings ?

At the moment I am using Dream kernel with Extweaks which offers a large amount of customisation. I can set the app to UV -75mv if i leave the gpu settings relatively untouched. With -50mv I seem to be able to play with the GPU settings without freezing as often.

I know trial and error will show the best results but there are a lot of variables to test which makes it time consuming.

Anyone know the relative benefits of each method ? Or is it too connected to break down this way.

 

[ironass]

When it comes to tweaking, everything is inter-connected.

The kernel you use, the ROM, the app's on the phone, your usage... all are inter-connected. Settings that will work for one person, will not be good for another.

I spent years trying to O/C, U/C, U/V not to mention the myriad of other settings that can come with some kernels... and at the end of the day, they saved so little juice and impaired performance as well as wasting so much of my time for very little or no return, that by-and-large, I stick with the developers settings.

Rule of thumb for me... "If the kernel and the developer are good enough, so are their settings!"

 

[robinta]

I can see your point, and I certainly can waste a lot of time for questionable benefits.

But I think a lot of kernels are there to be tweaked depending on whether you want out and out performance, stunning battery life or somewhere in between.

I've definately seen some improvements with different setups. Whether its a placebo effect I don't know. It also satisfies by urge to fiddle with settings !

 

[Slug]

Does undervolting (i.e applying a reduction to each frequency) or reducing the settings for the gpu result in the best battery savings ?

Yes. No. Maybe. It depends on who you believe, and personal experience.

In isolation, a (as an example) 50mV CPU undervolt reduces power consumption by very little. Likewise similar for the GPU. Put together however, along with reductions in clock frequencies, modified stepping tables, choice of scheduler and governor etc and an overall improvement can often be clearly seen. The trouble is that interconnection thing hinted at above.... the only way to find out what settings work best is to test, and that takes time and patience. And careful backups, 'cos you will freeze/hang/crash your handset during the process. I speak from bitter experience.