OverClocking 101

[tcyree2202]

Here is my "looking from 1 mile above" soap box.

There are a LOT of users that follow what their ROM makers say without doing any research. The best example of this would be Bugless Beast and Ultimate Droid ROMs. They are told in the release notes "You do not need to run SetCPU because overclocking is now built in". Well that is just fine and dandy, right up to you kill your phones battery life or worse kill the phones CPU while using the performance, interactive or a very poorly configured ondemand setup.

Now if everyone used SetCPU properly, then this wouldn't be an issue. Because they wouldn't make the mistake of not setting the sleep/standby profile to ondemand, conservative or powersave. But that isn't the case. Hence why this thread actually exists. Because the average person simply doesn't have time to do the research needed to figure out "WHY" you need a sleep/standby profile that isn't set to performance or interactive. Or better yet understand that there are certain governors you should "NEVER" mix because SetCPU doesn't maintain different entries for each of the fields that you can set and will cause issues that look like the phone is lagging out or even overheat.

Interactive is better than the performance governor. No IF, ANDs or BUTs about it. But you have to know WHEN and WHY to use it. And that isn't the average users forte. Hell, lets be realistic, I don't even know how the new conservative governor works yet because I simply don't have time to research it with my current work and home life. And that is with me already investing over 100 hours doing research, testing and code reading. One of these days I hopefully will get to it.

/soap box off


Short answer to your question is: Yes, when properly setup and with correct supporting profiles, the interactive governor is a great asset to the Android OS. But one mistake in the setup and it can eat your battery in less than 5 hours.

Thanks for the info and quick response! This whole rooting game is definitely a constant learning process.

 

[hrdnhvy]

I know this is probably a nob idea, but one thing that came to mind about underclocking while the phone is "sleeping" is do to with any apps that may run or update while ur phone is sleeping & The trade off to being underclocked vs the length of time it takes the app to complete the task, i mean being underclocked one would assume it would take slightly longer for the task to complete vs running at a higher speed one would assume it would take less time...what would affect the batter drain most...the cpu speed or the length of time the cpu is under the load.

 

[Skull One]

I know this is probably a nob idea, but one thing that came to mind about underclocking while the phone is "sleeping" is do to with any apps that may run or update while ur phone is sleeping & The trade off to being underclocked vs the length of time it takes the app to complete the task, i mean being underclocked one would assume it would take slightly longer for the task to complete vs running at a higher speed one would assume it would take less time...what would affect the batter drain most...the cpu speed or the length of time the cpu is under the load.

I have tested it both ways. I have seen power savings, especially with low voltage kernels, using around 400 Mhz for sleep/standby. The power savings though is based on EXACTLY the question you asked. The Android OS is constantly in motion. It has to be to be able to respond to outside events (SMS/MMS/Calls...) and internal software timers (GMail, EMail, Facebook...). But as you surmised there has to be a switch over point when a higher frequency is actually better for battery life. The problem is, I can't think of a good way to test for that switch over point. Which means for now, the best testing is done by you with your phone under normal use. Test one week at a low frequency and record how many times you had to charge. Then one week at the higher frequency. If the difference is less than a 15% savings, I would say go with the higher frequency. The phone will be more responsive coming out of sleep/standby and for some people that is a big deal.

I wish I could give better answers than general theory. But because there are now over 100 different Android phones, 100K Android Apps and who knows how many different ROMs and kernel combos, theory is the only constant.

 

[hrdnhvy]

Kewl man & thanks for the reply....much like how kernels are picky on a phone to phone basis I would assume so would the above, some phones may work better one way & some phones the other, similar possibly to the ulv kernels work for some phones & others not, not all phones can be created 100% equal with the internals one phone may have a litrle more "resistance" to power consumption while others dont. These are the type of questions I come up when Im bored at work & "over-analyze" all the possible variables. Thanks again.

 

[moset]

I had trouble with what appeared to be the touch screen triggering itself like typing random letters I'm texts and jumping around in apps as if I pressed selections but I was not touching anything. I am awaiting a warranty replacement. Since then I pulled the SD card and the problem seemed to disappear. I reformatted the SD and reflashed the stock OS and have not seen the problem again.
My question is, knowing that different phones have different characteristics and my old one has never been dropped and handles LV kernels but overheard with ULV what is the easiest way to compare the replacement to determine if it is better than my old one.
I would hate to give up a better phone for a lessor.
I could root the replacement and test it, maybe using Quadrant as a gauge.
Any thoughts?

Sent from my Droid using DroidForums App

 

[furbearingmammal]

Drop the fastest, lowest-voltage kernel your current phone can run on your replacement and test it overnight with heavy use. If it works fine then you know you have an equal, as long as there are no other issues like sticker bubbles on the keyboard or wonky sounds on the headphone jack.

Quadrant is so variable a source I wouldn't rely on it for anything but bragging rights.

 

[Skull One]

I had trouble with what appeared to be the touch screen triggering itself like typing random letters I'm texts and jumping around in apps as if I pressed selections but I was not touching anything. I am awaiting a warranty replacement. Since then I pulled the SD card and the problem seemed to disappear. I reformatted the SD and reflashed the stock OS and have not seen the problem again.
My question is, knowing that different phones have different characteristics and my old one has never been dropped and handles LV kernels but overheard with ULV what is the easiest way to compare the replacement to determine if it is better than my old one.
I would hate to give up a better phone for a lessor.
I could root the replacement and test it, maybe using Quadrant as a gauge.
Any thoughts?

Sent from my Droid using DroidForums App

Your phantom inputs were caused by a bad kernel, not a bad phone. Returning to stock voltage, based on your description, cleared the issue. So if you are 100% certain that your phone can handle a LV kernel and there are no other issues with the phone, it is always going to be better than a refurb. The reason that a phone becomes a refurb is that it had a problem. I doubt they fix every problem perfectly. So a known good phone is always better than an unknown refurb in my opinion.

 

[hrdnhvy]

This comment/question is relation to my other comment/question that I made in post #499. Now I dont know how much of a factor in any it would be but its just another variable I thought of o add to my question in post #499.

In regards to clocking the speed of ones phone down while it is "sleeping" Now I would assume 99% of people would do that using the profile option in Setcpu. I have noticed while doing so, it obviously has to keep a running process going (which can be seen in running apps) in order for it to do so, so would or could that running process be a variable as well that may contribute I would imagine, a very slight amount, but none the less its running, contribute to battery consumption as well versus not having a profile set up. For me, while I am at work I am only able to check m phone every 2-4 hours anyway, so what I have been experimenting with is instead of setting up a profile, I simply keep a shortcut on my main screen to Setcpu & when I know I wont be accessing my phone for several hours I simply open Setcpu & turn down the max clock speed = no profile process runs.

I am going to try to figure out a way to test battery life doing that (manually setting down the clock speed when not in use to probably 400) compared to just leaving the phone clocked to 800 all of the time to attempt to compare things as I stated in post #499.

 

[moset]

Your phantom inputs were caused by a bad kernel, not a bad phone. Returning to stock voltage, based on your description, cleared the issue. So if you are 100% certain that your phone can handle a LV kernel and there are no other issues with the phone, it is always going to be better than a refurb. The reason that a phone becomes a refurb is that it had a problem. I doubt they fix every problem perfectly. So a known good phone is always better than an unknown refurb in my opinion.[/QUOTE]

I don't think the kernel or the ROM was to blame. I reverted to stock and the problem persisted. After removing the SD card the problem disappeared. I reinserted the SD and it wouldn't mount. I put another SD in and no problem. I then formated the original and reloaded the stock OS and for 24 hrs have had no problem.
The original SD could be read on my computer but not the phone until I formatted with the phone.
I agree about the referb being a referb because someone else had trouble. If I'm lucky I could get one that was returned for the 30 day worry free deal. Wish there was some kind of diagnostic I could run on both.
Skull, thanks for your advice. I have read ur stuff and appreciate your research.

Sent from my Droid using DroidForums App

 

[Skull One]

Your phantom inputs were caused by a bad kernel, not a bad phone. Returning to stock voltage, based on your description, cleared the issue. So if you are 100% certain that your phone can handle a LV kernel and there are no other issues with the phone, it is always going to be better than a refurb. The reason that a phone becomes a refurb is that it had a problem. I doubt they fix every problem perfectly. So a known good phone is always better than an unknown refurb in my opinion.

I don't think the kernel or the ROM was to blame. I reverted to stock and the problem persisted. After removing the SD card the problem disappeared. I reinserted the SD and it wouldn't mount. I put another SD in and no problem. I then formated the original and reloaded the stock OS and for 24 hrs have had no problem.
The original SD could be read on my computer but not the phone until I formatted with the phone.
I agree about the referb being a referb because someone else had trouble. If I'm lucky I could get one that was returned for the 30 day worry free deal. Wish there was some kind of diagnostic I could run on both.
Skull, thanks for your advice. I have read ur stuff and appreciate your research.

Sent from my Droid using DroidForums App

My bad. I re-read your original post and see where I missed the SD card being the key.

LOL. Guess I am too used to blaming bad voltages for everything

 

[Skull One]

This comment/question is relation to my other comment/question that I made in post #499. Now I dont know how much of a factor in any it would be but its just another variable I thought of o add to my question in post #499.

In regards to clocking the speed of ones phone down while it is "sleeping" Now I would assume 99% of people would do that using the profile option in Setcpu. I have noticed while doing so, it obviously has to keep a running process going (which can be seen in running apps) in order for it to do so, so would or could that running process be a variable as well that may contribute I would imagine, a very slight amount, but none the less its running, contribute to battery consumption as well versus not having a profile set up. For me, while I am at work I am only able to check m phone every 2-4 hours anyway, so what I have been experimenting with is instead of setting up a profile, I simply keep a shortcut on my main screen to Setcpu & when I know I wont be accessing my phone for several hours I simply open Setcpu & turn down the max clock speed = no profile process runs.

I am going to try to figure out a way to test battery life doing that (manually setting down the clock speed when not in use to probably 400) compared to just leaving the phone clocked to 800 all of the time to attempt to compare things as I stated in post #499.

SetCPU has "Triggers" setup that actually wake it up for 90% of what it does. I.E. every profile but heat is triggered by an event and hence requires no running code while waiting for that event to fire off. Now heat does have to be sampled. Based on observations I have done, I would say around every 10 minutes while asleep it samples the temps and at least every 5 minutes when in use. That means the program has to wake up long enough to sample, process and then go back to sleep. I can't see that taking longer than 3/4 of a second to maybe 2 seconds if there is a lot going on.

 

[aggiechase37]

Come on Skull! You've got to look into the conservative governor! I'm telling you now! It's what laptop linux people use to conserve battery. With regular use I'm on hour 26 right now with 40% left.

CM 6.1 rc3 by the way.

Come on everyone, sing it with me! Conservative governors! Conservative governors!

 

[aggiechase37]

30 hours of normal use with conservative. I am now plugging in. I think I could have gotten another hour. setcpu says I have 10% left. Just sayin!

 

[Skull One]

Come on Skull! You've got to look into the conservative governor! I'm telling you now! It's what laptop linux people use to conserve battery. With regular use I'm on hour 26 right now with 40% left.

CM 6.1 rc3 by the way.

Come on everyone, sing it with me! Conservative governors! Conservative governors!

I will do a NON-scientific test and run conservative for the next two weeks. I will adjust the advanced stuff every 2 or 3 days to see how it affects things so that I can hone in on a decent overall setting. I have charged the phone to 100%, deleted the battery stats and rebooted. When I hit 20% power I will create a copy of the time-in-state file so I can track how often each slot was in use. I will also note how much display time was used.

This isn't my normal methodology but I should be able to glean enough information to make some suppositions.

 

[aggiechase37]

You're going to be surprised I think. Can't wait for your analysis!

Sent from my Droid using DroidForums App