OverClocking 101

[NetNathan]

Nice thread.
Good explaination.


Sent from NetNathan's Droid using Tapatalk.

 

[Skull One]

Froyo 2.2 Beta

Froyo 2.2 Beta.

I have been running it for 24 hours now. Of course the overclocking part of me set out very quickly to find out what can be done.

First thing I noticed different was the base line ondemand settings.
2.1 defaults to 48650 (or something close to that) , 86, 0, 0.
2.2 defaults to 30000, 93, 0, 0.

That change is actually pretty significant in my opinion. Instead of checking to see if a frequency change is needed 21 times a second, it now checks 33 times a second. That is a 60% increase in the number of checks per second. So my first question was "What does that affect overall"? I can tell you this much based on my current observations and testing, it doesn't hurt battery life in any way I can find.

The benchmarks are where I have seen the biggest differences. With a 1Ghz kernel Linpack would top out at 8.4 MFLOPS on 2.1. Always had to run the benchmark 5 times in a row to get an idea what the top end was. Mainly because the first run would be in the 5.5 to 6.2 MFLOP range and it would increase till the 4th or 5th run. I have assumed it was due to the OS loading the APK fully into memory, putting other tasks to sleep and switching its complete focus to app asking for the most CPU time. What ever it was, it seems to be gone in 2.2. Linkpack now show 15.8 MFLOPS with the first runs always at 14.x. And this is just a test build for the Droid. It isn't even a release candidate. Hopefully the final build will be even faster.

Battery life is much improved based on my current subjective testing. Sleep/Standby is VERY MUCH improved. When I wake the phone up, it is snappier and the load avg never reads above 1.5. Under 2.1 after just 3 hours of sleep mode the phone would be a tad slow and the load avg would read in the high 80s.

And my big complaint "may" finally be fixed. So far every phone call with a custom ringtone has worked correctly. I have never had that happen with 2.1. Now this is still early, so it could fail me eventually, but for now I am hopeful.

I plan on staying with the Froyo 2.2 Beta until a show stopping bug bites me or a newer version is released. Because this version of the OS truly shows what the Motorola Droid is capable of.

 

[pc747]

Froyo 2.2 Beta.



I plan on staying with the Froyo 2.2 Beta until a show stopping bug bites me or a newer version is released. Because this version of the OS truly shows what the Motorola Droid is capable of.

+1 on that, atleast until the official version with root is released.

 

[mkthnx001]

So are you saying running something like 62500, 70, 0, 70 wouldn't be suggested anymore on 2.2?

 

[jroc]

^ I was gonna ask this too. Can we change the Advanced settings to what some of us used on 2.1, and will it make any differences?

 

[Aixelsyd]

@Skull One nice writeup. i've been using Froyo on a spare Droid for a while now and totally agree with your findings. I will say this, the newer builds of Froyo are even better from what I've seen. This version going around it really not to new and a bit old to be perfectly honest. The more current version is even better that is for sure.

 

[Skull One]

So are you saying running something like 62500, 70, 0, 70 wouldn't be suggested anymore on 2.2?

I am testing 30000,93,0,0. Pulled the Droid off the charger at 12:03 am. I am still at 70%. I am in normal usage mode today. Rough guess I will be at 60% battery instead of my normal 40% in 24 hours. And I haven't even installed the new base band yet to see what that does.

I am probably going to test 50000, 75, 0, 100 for tomorrows run. And then install the base band Wednesday and run those same numbers. I have a feeling I might get 36 hours out of the battery.

 

[mkthnx001]

So are you saying running something like 62500, 70, 0, 70 wouldn't be suggested anymore on 2.2?

I am testing 30000,93,0,0. Pulled the Droid off the charger at 12:03 am. I am still at 70%. I am in normal usage mode today. Rough guess I will be at 60% battery instead of my normal 40% in 24 hours. And I haven't even installed the new base band yet to see what that does.

I am probably going to test 50000, 75, 0, 100 for tomorrows run. And then install the base band Wednesday and run those same numbers. I have a feeling I might get 36 hours out of the battery.

Haha, 40% in 24 hours? That's insane in my point of view. I don't know if it's because I send and receive about 400 text messages a day or what, but I wouldn't dream of that. You wouldn't be using an extended battery would you?

Also, are you saying you need to do further testing before you decide whether a sampling rate of 30000, etc., is better than 62500 (ex.)?

 

[Skull One]

Haha, 40% in 24 hours? That's insane in my point of view. I don't know if it's because I send and receive about 400 text messages a day or what, but I wouldn't dream of that. You wouldn't be using an extended battery would you?

Also, are you saying you need to do further testing before you decide whether a sampling rate of 30000, etc., is better than 62500 (ex.)?

I am using the stock battery.

30000 over 62500 is an interesting technical question.

It provides faster frequency switching in both directions. Which means it can stay at 250Mhz (if that is where you have the low set) and then can very quickly jump to your highest frequency in 3 samples (or 90 milliseconds). Which means the phone will be very responsive. The down side to that, the phone is using twice the CPU time to check that status. But that can be compensated for by the OS coding. Which is what I think is happening now. I believe 2.2 is far more optimized than Google is letting on. We won't know for sure till we get to look at the source. But I have my hunches.

The very fact that sleep/standby in 2.2 now works with my custom ringtones and Voice Caller ID with no issues, shows me they made major changes to the core part of the OS. I am now at 20 phone calls with everything working properly. 2.1 never made it past 20 no matter what tricks I tried.

 

[mkthnx001]

Alright, I look forward to reading about your future tests.

Also, do you do anything special to have battery life like that? Do you send only a few text messages a day, or hundreds? Texting can be a major battery drain, right?

Awaiting your response.

 

[Skull One]

Alright, I look forward to reading about your future tests.

Also, do you do anything special to have battery life like that? Do you send only a few text messages a day, or hundreds? Texting can be a major battery drain, right?

Awaiting your response.

Texting uses two major battery drainers. Display and 2G/3G. Why Verizon hasn't asked Motorola to detect WiFi and send texts over it is beyond me. The battery savings would be around 50% over 2G/3G based on some testing I did.

In fact, here are some interesting numbers from OTA 2.1:

Phone was setup to be as dumb as possible. I erased every app I could so nothing could auto launch or background load. I set the display to 100% brightness and clocked the CPU at 250Mhz. I full charged the battery to 100% and then let it drain to 80%. I recorded the number of minutes between 90% to 80%.

100% brightness @ 250 Mhz. 10% battery drain. 65 minutes.
Wifi download with above, 25 minutes.
3G download with above, 16 minutes.

Now the above numbers above are technically useless as a whole, which is why I stopped the testing, for true power usage comparison. The reason was simple. The first test didn't have any writing to the SD Card. Now the second and third test did, so they can be compared against each other. WiFi is 56% more efficient than 3G when it comes to battery life. But what I can't tell you is how much of that test is 3G power vs display power.

My hunch is the display time is eating more power than your text messages actually being sent. One day I will figure out a way to test with a level playing field that doesn't require me to write any code. I prefer to use built in apps only for the tests.

 

[mkthnx001]

I thought texting didn't require 3G? As far as Juice Defender is concerned anyway. Although after observing it, I think it does. That would explain a lot.

Also, like I said, how does your battery last that long? Do you not text at all, or rarely?

 

[Skull One]

I thought texting didn't require 3G? As far as Juice Defender is concerned anyway. Although after observing it, I think it does. That would explain a lot.

Also, like I said, how does your battery last that long? Do you not text at all, or rarely?

I prefer to talk over texting. So maybe 20 a day because of my children. And of those 2 or so are pictures of my grand kids doing something funny.

 

[mkthnx001]

It just blows my mind that you get over 24 hours of battery life.

On a normal day like today I send/receive about 200 texts, talk on the phone for about 30 minutes, browse the web, and listen to music for about 30 minutes each as well.

All of this gets me about 6 hours of battery life.

Anything you do different?


EDIT: I forgot to mention I run the following under SetCPU:

Default: 700MHz/250MHz
Battery < 51%: 600/250
< 31%: 550/250
Sleep/Standby: 300/250

Advanced: 62500, 70, 0, 70

 

[Skull One]

Froyo 2.2 part 2

Froyo 2.2 part 2

I ran for 24 hours with SetCPU basically in a default mode: 250/1000, 30000, 93, 0, 0. I was hoping to be in 60% battery range at the end of that test. Alas it didn't come to pass. I tripped over to 40% at the 23 hour mark. So I started recharging and reset the phone to: 250/1000, 50000, 75, 0, 100. Since I normally wake up at least once at night, I unplugged and made sure the test was properly setup. Four hours later, I awoke and found the phone very sluggish. So I checked the standard programs and found something odd:

SetCPU was reporting a load avg of 3.9x. Prior 24 hour test never was above .8 while asleep/standby.

Then I checked System Panel and the CPU never got to 1000 Mhz. Kept topping out at 900.

So I checked my Sleep/Standby setting and saw that I had set it to 250/400. On 2.1, that setting NEVER made a difference in the world. But in 2.2 it makes ALL the difference. 2.2 is actually respecting SetCPU's Sleep/Standby with OnDemand settings and idled all night at 250Mhz because the power bias never let it get to 400 mhz to play catch up.

So the power bias setting seems to be more powerful than it was under 2.1. It will be fun to test just how much it can affect the system.

But here is where I am getting really excited. 125Mhz. I believe 2.2 just might be able to handle it because if the CPU is truly switching frequency while asleep and 125/400 should keep the Droid caught up and still allow the system to wake up properly for texts and phone calls.

Now, who do I beg to for a 125/250/400/600/700/800/1000 Low Voltage kernel for 2.2?