[Updated 11-Dec-2018][oxygen] Battery Life / Performance Benchmarks - AOSP Roms

[pauldrudd]

That sounds like a great idea! Would love to see MSM with Hardrock Kernel

 

[911-Future_Maker]

Excluding the first result Msm got 13h53m sot on average and RRO 14h52m both on overclocked kernels. They should pass 16h on normal 2ghz frequency making them best roms for battery ever. :good: @svbarbosa can you do only one test of msm with schedutil governer and reduced cpu clock to normal 2ghz? Many thanks for the time you spend making these tests, they're very useful.

 

[Shaolian]

Hi, I'm currently using Jay_B's DotOS (Nougat) rom.

How do I change CPU Governor and CPU Hotplug? I've install "Kernel Adiutor", but couldn't find any settings for "Hotplug". What I can see are "LITTLE governors" and "big governors".

Reasons are that I want to recreate the settings that were used for this performance benchmark.

Thanks in advance. Sorry if this hass been asked before.

 

[yepyup]

good information thx

 

[Jay_B]

Hi, I'm currently using Jay_B's DotOS (Nougat) rom.

How do I change CPU Governor and CPU Hotplug? I've install "Kernel Adiutor", but couldn't find any settings for "Hotplug". What I can see are "LITTLE governors" and "big governors".

Reasons are that I want to recreate the settings that were used for this performance benchmark.

Thanks in advance. Sorry if this hass been asked before.

Are you on Hardrock Kernel for nougat based ROMs. If yes then then only try those changes in cpu or gpu.

 

[ayush4261]

Hi, I'm currently using Jay_B's DotOS (Nougat) rom.

How do I change CPU Governor and CPU Hotplug? I've install "Kernel Adiutor", but couldn't find any settings for "Hotplug". What I can see are "LITTLE governors" and "big governors".

Reasons are that I want to recreate the settings that were used for this performance benchmark.

Thanks in advance. Sorry if this hass been asked before.

Hot plug is not included in hard Rock kernel. I requested the same but he clarified it's not needed for our device.

 

[svbarbosa]

Good News and bad news.

I found out the reason of the large SOT variances in my benchmarks.

You are probably aware that I start every benchmark with battery at 80%.

Have you ever wondered how I achieve this? Simple - I use an app that stops charging at an specified percentage.

The app I use and recommend is [APP][ROOT][4.0+]Battery Charge Limit.

This is a tricky app. The way to control charging varies a lot between devices.

Somehow, for the Max 2, the way the app chooses by default to control charging seems to work, but does not!!

Yes, charging LED turns off and yes, battery percentage indicator stays at 80%, but the battery keeps charging.

I figured this out after one of my runs with Hardrock kernel 13.1 reached 22 hours SOT with extreme overclocking. This simply could not be.

After this result I started checking battery voltage levels and found the bug.

Did some research and tests and was able to configure BCL to properly stop the charging at desired level.

All you have to do is change the control file to "qpnp-smbcharger-18" and voilá, all is good again.

The bad new is that the many dozens of benchmarks I have run so far may not be trusted.

The problem only affects the SOT, of course. Performance scores are good regardless of the battery level problem.

I found some ways to figure out bad benchmarks and filtered the ones I ran for Oreo roms. This is my last preview adjusted:

This explains why @Jay_B´s MSM Xtended got the 20 hours SOT. Unfortunately a false result.

Anyway all results from now on will be free from this inconsistency.

I was in the middle of testing kernels using the same rom. RRO 2018-0701.

Still have to decide If I will try to save all good benchmarks from this series or start over again.

My kids school vacation is over this weekend. At least I will have more time from now on.

 

[ne0t]

TBH I have no desire of using MIUI, but since you are asking nicely and @jbmc83 endorses it I will do.

Anyway MIUI will probably have a good SOT since it kills background processes like crazy.

Which MIUI build should I benchmark?

First, thanks a lot for your hardwork doing these benchmarking on Mi Max 2's ROMs (can't imagine how you did it..)..

About MIUI ROM, basically there are 2 branches, Stable ROM and Developer ROM (beta) :

*Latest Developer Global ROM (MIUI 10) * :
Recovery 1611M :
http://bigota.d.miui.com/8.7.26/miui_MIMAX2Global_8.7.26_8966518c5b_7.1.zip
Fastboot 2121M :
http://bigota.d.miui.com/8.7.26/oxy...26_20180726.0000.00_7.1_global_98bc71a4c1.tgz

*Latest Stable Global ROM (MIUI 9) * :
Recovery 1538M :
http://bigota.d.miui.com/V9.5.4.0.NDDMIFA/miui_MIMAX2Global_V9.5.4.0.NDDMIFA_fecf400643_7.1.zip
Fastboot 2087M :
http://bigota.d.miui.com/V9.5.4.0.N...FA_20180406.0000.00_7.1_global_fc8db01904.tgz

Hope you could find some time doing benchmark on these ROMs, thanks !

 

[Horned Duckie]

Good News and bad news.

I found out the reason of the large SOT variances in my benchmarks.

You are probably aware that I start every benchmark with battery at 80%.

Wouldn't it be easier and, in fact, faster just to run all those battery life tests with battery charged 100%? ?

 

[svbarbosa]

Wouldn't it be easier and, in fact, faster just to run all those battery life tests with battery charged 100%? ?

First 100% percentage may also be wrong. Some phones will keep charging the battery even after reaching 100%. So starting at 100% is not error proof.

I have added an Battery Voltage monitor app to my bag of tools. It displays the voltage in status bar, so I can always check it and it goes in all my screenshots. I believe this is the best way to make sure the battery percentage is correct, or at least the voltage is the same.

For the Mi Max 2, with the huge battery, the PC Mark battery test takes a VERY LONG time.

Say rom X has an SOT of 18 hours. If I start the benchmark at 100% it will run until charge drops below 20%, so it will take 14 hours and 24 minutes to run. With this duration I will only be able to run one benchmark per day.

Starting with battery at 80%, the benchmark will take 10 hours and 48 minutes, increasing the chances I run two benchmarks in one day.

 

[svbarbosa]

Here is a preview of kernels benchmarks.

The second hardrock set of benchmarks was done with the battery percentage problem solved.

For previous benchmarks I had to filter out bad benchmarks.

Now the SOT variances are back to "normal below 1%" values. :victory:

I am right now benchmarking Xtended x9.5, then I will go back to kernels with a fresh start.

 

[jbmc83]

Here is a preview of kernels benchmarks.

The second hardrock set of benchmarks was done with the battery percentage problem solved.

For previous benchmarks I had to filter out bad benchmarks.

Now the SOT variances are back to "normal below 1%" values. :victory:

I am right now benchmarking Xtended x9.5, then I will go back to kernels with a fresh start.

hey man, glad uve sorted out the troubles with the SOT variances! kudos to your perseverence man! this is a LOT of work! i just hope ure also having a bit of fun doing this

cheers mate!

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

[svbarbosa]

hey man, glad uve sorted out the troubles with the SOT variances! kudos to your perseverence man! this is a LOT of work! i just hope ure also having a bit of fun doing this

cheers mate!

Thanks mate. :good:

Time flies. I started benchmarking Mi Max 2 roms in mid april, almost four months ago!!

TBH the first weeks were very frustrating. Benchmarks would not finish and when they did most results were inconsistent.

I should have suspected there was something wrong with battery percentages earlier. I had some hints before but overlooked them. :silly:

At least now it is all sorted out. Too bad I have practically to start all over again.

But I am resilient. The show must go on.

 

[Shaolian]

Are you on Hardrock Kernel for nougat based ROMs. If yes then then only try those changes in cpu or gpu.

Hot plug is not included in hard Rock kernel. I requested the same but he clarified it's not needed for our device.

Hi, thanks for the replies, but I only understand now after flashing Hard Rock kernel that the Hotplug option isn't available. But no matter, the overclocked CPU to 2800 MHz and GPU to 1300 MHz is working flawlessly.

It's just amazing how the Mi Max 2 can be overclocked without any overheating issues :good:

 

[svbarbosa]

Here is another series of kernel benchmarks.

This time MSM Xtended is the rom used to test kernels.

Special thanks to @Jay_B, @Martichou and @Hard_Rock83 for supporting so well the Mi Max 2.

Finally benchmarks may be really trusted. All these were done with the battery percentage problem solved.

AceKernel is incredibly stable. Three runs with the same exact SOT. Two of these with exactly the same performance score. This is a first after running hundreds of benchmarks. :good:

I have an explanation about how I tested hardrock kernel:

Forgive me @Hard_Rock83, but I am not brave enough to run benchmarks for hours (I really mean hours, each benchmark takes at least 10 hours) with your extreme overclock.

I know a lot of people are using your OC kernel with no problems at all, but phones and chipsets are not created equal and tolerances to OC may and will vary between same model phones. Maybe only one out of every thousand will suffer with OC but I would hate to be that one.

That being said, what I did was to apply Balanced profile from Hardrock_kernel_profilesv4.zip. The profile took care of GPU max frequency. I changed CPU min/max frequencies to 480/2016 to disable CPU OC.

@jbmc83, I am willing to benchmark your fine tunning of hardrock kernel if you like.

Please PM me or post here your complete current settings, or better yet, create a KA profile and send it to me.

 

[Jay_B]

Thank you bro for the time and patience. Appreciate the details here. Good work .

 

[jbmc83]

Here is another series of kernel benchmarks.

This time MSM Xtended is the rom used to test kernels.

Special thanks to @Jay_B, @Martichou and @Hard_Rock83 for supporting so well the Mi Max 2.

Finally benchmarks may be really trusted. All these were done with the battery percentage problem solved.

AceKernel is incredibly stable. Three runs with the same exact SOT. Two of these with exactly the same performance score. This is a first after running hundreds of benchmarks. :good:

I have an explanation about how I tested hardrock kernel:

Forgive me @Hard_Rock83, but I am not brave enough to run benchmarks for hours (I really mean hours, each benchmark takes at least 10 hours) with your extreme overclock.

I know a lot of people are using your OC kernel with no problems at all, but phones and chipsets are not created equal and tolerances to OC may and will vary between same model phones. Maybe only one out of every thousand will suffer with OC but I would hate to be that one. [emoji14]

That being said, what I did was to apply battery profile from Hardrock_kernel_profilesv4.zip. The profile took care of GPU max frequency. I changed CPU min/max frequencies to 480/2016 to disable CPU OC.

@jbmc83, I am willing to benchmark your fine tunning of hardrock kernel if you like.

Please PM me or post here your complete current settings, or better yet, create a KA profile and send it to me.

my man! incredible work very impressive indeed.

thought id make ur life a bit easier and create a profile for you, so no need to implement all settings by hand one by one into kernel adiutor

Link: https://mega.nz/#!0H5EGYLY!KBbXBna6vmFBApSQXy1wkohyTx5SkVthPPBitChh9-A

im guessing the scores will be much lower on this profile since i tried to optimize it for battery life without sacrificing subjective performance during day to day operation. but very curious to see how much, if at all, battery life will profit

cheers mate!

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

[svbarbosa]

...
Thought id make ur life a bit easier and create a profile for you, so no need to implement all settings by hand one by one into kernel adiutor

Link: https://mega.nz/#!0H5EGYLY!KBbXBna6vmFBApSQXy1wkohyTx5SkVthPPBitChh9-A

im guessing the scores will be much lower on this profile since i tried to optimize it for battery life without sacrificing subjective performance during day to day operation. but very curious to see how much, if at all, battery life will profit
...

Guess you're anxious to see results. Here are screenshots of the first two runs.

 

[jbmc83]

Guess you're anxious to see results. Here are screenshots of the first two runs.

haha no stress, take all the time you need!

(but yeah, im maybe a little bit curious )

interesting, so work score is in the range of 75-85% of the OREO roms uve tested so far, with battery life 10 to 13% higher (about 1.5h SOT extra). funny, i expected lower scores but higher battery life gains

 

[svbarbosa]

haha no stress, take all the time you need!

(but yeah, im maybe a little bit curious )

interesting, so work score is in the range of 75-85% of the OREO roms uve tested so far, with battery life 10 to 13% higher (about 1.5h SOT extra). funny, i expected lower scores but higher battery life gains

The PC Mark SOT is calculated for 95% of battery usage, not 100%, so a 16 hours SOT translates into 16 hours and 50 minutes for 100% battery.

Even with this adjustment, I have a feeling that these PC Mark SOT figures tend to be smaller than what you would get in real world.

PC Mark tries to emulate normal usage with many tasks, but looking at its screens while its running, I get the impression that it is somewhat accelerated.

When a real person is performing the same tasks, he/she will always be at a slower pace. Sometimes taking a look around, or talking to someone with the screen on waiting, or giving small pauses for many other reasons. The battery has a lot of "rest moments" and drain will be smaller during these moments. PC Mark imposes a higher pace with no "rest moments".

So a real person usage will most certainly translate into a higher SOT than the PC Mark benchmark, even after adjusting the SOT to 100% battery.