FORUMS
Remove All Ads from XDA

Haldi's Benchmark Thread for Governors&Undervolting&Overclocking the Z2

2,322 posts
Thanks Meter: 785
 
By Haldi4803, Senior Member on 1st June 2014, 02:05 AM
Post Reply Subscribe to Thread Email Thread
1st October 2015, 10:59 PM |#141  
Haldi4803's Avatar
OP Senior Member
Thanks Meter: 785
 
More
Another Guest Appearance: Xperia Z
My old Sony Xperia Z is still being used by other family members.
So i took the chance when it was empty and charged it full Using the Aukey PA-U28 QuickCharge2.0 Charge i use for my Z2.




The Old 2330mAh battery in the Z only goes from 3.8 to 4.2Volts and already reaches at 65% Charged the maximum Voltage. This happens after 70 Minutes, it takes another 80 Minutes to fully charge to 100%
 
 
2nd October 2015, 04:32 PM |#142  
Senior Member
Thanks Meter: 211
 
Donate to Me
More
So @Haldi4803, we're getting closer to a really safe QC2.0, what do you think we can improve more? Or is the 2000 mAh kernel safe enaugh?

Sony Xperia Z2
13th October 2015, 12:02 AM |#143  
Haldi4803's Avatar
OP Senior Member
Thanks Meter: 785
 
More
Off Topic:
Insides of the Z2


So... yeah... i've managed to break the back glass of my Z2.
After dozens of crashes on the floor one unlucky day my back glass shattered.

Well, Shít Happens, not that much of a problem. Bought an 8$ replacement back cover from Witrigs, while at it i've also bought a new Headphone Jack (my old was having problem getting a stable connection) and a Tempered Glass Screen Protector (well, if it protects the front it will also work on the back, paid 25$ express shipping so another 11$ for the Screen Protector aren't that expensive^^)



Replacing a broken back cover is quite a hassle. It shatters more and more ^^



if only replacing the Back cover i'd be finish at this point, but yeah, the headphone jack.... hidden under the mainboard. yay.



Here you can see the Mainboard and the heatpipe for the SoC (jep the SoC is also covered by a heatspreader lame can't see stuff )



Headphone jack is gone (that stuff is waterproof and glued on... damn... realy good glue there)



so yeah, i've put everything back together and cleaned the borders for the new back cover.



Phone is back and everything is working, no problems with overheating and stuff, everything like always.




There are just two little but very important points to know when doing this.
A: You cannot unmount the mainboard when a microSD card inserted, it's physically not possible the card is to big and blocks.
B: When using a heater to melt the glue on the back cover, make so to unplug the microSD card before, these damn suckers don't like heat.

Oh yeah, and the Backside foil from the Nillkin Tempered Glass Screen Protector is a simple plastic foil and not the same 0.2mm tempered glass like on the Frontside -.-
The Following 6 Users Say Thank You to Haldi4803 For This Useful Post: [ View ] Gift Haldi4803 Ad-Free
30th November 2015, 05:38 PM |#144  
Haldi4803's Avatar
OP Senior Member
Thanks Meter: 785
 
More
Powerdrain per Mhz
Is overclocking worth it? How much will you save by underclocking?

If you have taken an interest in this thread you most likely will have read some Smartphone Reviews on Anandtech. I really like them because they know what they're talking about, imho they're the most knowledgeable website out there for in dept reviews of how a SoC really works.
Seeing this chart (review S6)which shows Power Consumption per Mhz Frequency i was impressed and tought we could use something similar for our Z2.

So one day i had to much free time and used Qualcomm Trepn Profiler, StabilityTest v2.7 and Kernel Adiutor to start benching our Z2.
Because the "User" Hotplug does not exist, or not work with Kernel Adiutor, i used Alucard Hotplug and set the minimum and maximum active Cores.
To aquire similar conditions for all benchmarks the phone was put in Flightmode without WiFi, Manual Brightness at Minimum.
Benchmarking for 30 seconds each frequency gave an average Power usage.



The Result is this horrible looking table.
Code:
Cores/ mhz 268	300	422	652	729	883	960	1'036	1'190	1'267	1'497	1'574	1'728	1'958	2'265	2'457	2'572	2'726
4	1'592	1'653	1'885	2'089	2'259	2'401	2'661	2'701	2'893	3'089	3'445	3'669	3'940	4'454	5'091	5'486	6'492	6'621
3	1'384	1'426	1'670	1'833	1'945	2'036	2'238	2'332	2'500	2'633	2'884	3'069	3'150	3'812	4'297	4'812	5'046	5'433
2	1'142	1'238	1'391	1'559	1'636	1'745	1'869	1'929	2'030	2'182	2'401	2'501	2'682	3'023	3'257	3'540	3'787	4'110
1	903	905	1'007	1'123	1'233	1'250	1'474	1'550	1'544	1'679	1'740	1'846	2'048	2'046	2'320	2'368	2'571	2'536
And this beautiful Graph


Now it's your choice to underclock to 1728mhz instead of 1958mhz because the gain is worth it
But be aware that all this benchmarks are made with full CPU load, don't forget the "race to idle" principle when underclocking to much.
Last edited by Haldi4803; 30th November 2015 at 05:44 PM.
The Following 5 Users Say Thank You to Haldi4803 For This Useful Post: [ View ] Gift Haldi4803 Ad-Free
10th December 2015, 11:01 PM |#145  
Haldi4803's Avatar
OP Senior Member
Thanks Meter: 785
 
More
Another Guest Appearance: Samsuns Galaxy S6 Edge
So, i got my Hands on a Samsung Galaxy S6 Edge with 2600mAh battery.
Installed the required apps and charged it from 2% to 100% only 80 Minutes, extremly fast, i'm impressed.


25th December 2015, 11:35 AM |#146  
Mokiwipeout's Avatar
Senior Member
Thanks Meter: 83
 
More
Quote:
Originally Posted by Haldi4803

Powerdrain per Mhz
Is overclocking worth it? How much will you save by underclocking?

And this beautiful Graph


Now it's your choice to underclock to 1728mhz instead of 1958mhz because the gain is worth it
But be aware that all this benchmarks are made with full CPU load, don't forget the "race to idle" principle when underclocking to much.

Do you think that underclocking all the way down to 1.19 GHz is actually worth the gain(given that the phone spends most of screen on time in this frequency) or that 1.7 GHz is a better option given the fact that consumption increases by only 25% or something
25th December 2015, 03:40 PM |#147  
Haldi4803's Avatar
OP Senior Member
Thanks Meter: 785
 
More
To accurately answer this question we need to take a closer look at the performance on this frequency.
If your phone takes longer to solve your task the powerdrain per Task might not be as great on the low frequencies.

Theoretically it would be possible to get the performance in MIPS for a certain frequency and then take the powerdrain from the graph to find the perfect efficiency.

Sadly that's purely theoretical performance and does not necessarily reflect real world usage.

Might be possible to find out by a real-world Task like picture editing or other stuff that has a 30-90second duration and shows a clear start&end point and use Trepn profiler to measure powerdrain for all different frequencies.
I don't know of any task that would be doing that. Anyone has some ideas?
13th February 2016, 07:59 AM |#148  
Member
Thanks Meter: 21
 
More
i have one question, in the list of tests theres a test named "Undervolting a big hoax" and another one named "Undervolting, it does help"

yes, i have read both and in the second the clear conclusion was that undervolting will bring 17% battery increase so that cannot be a hoax (also the temp was a lot lower which is great)... so what am i missing here, is it that undervolting when scaling on the processor is enabled will not net any significant gain as opposed to static processor speed? i guess this could be the case since there are a lot of different factors at play whit processor scaling... any clarification would be welcome
13th February 2016, 08:31 PM |#149  
Haldi4803's Avatar
OP Senior Member
Thanks Meter: 785
 
More
No,
The difference lies in the usage.
The first Benchmark is made with the "normal usage" like any sane person would use a Smartphone, browse Web, read emails stuff like that.
You see 25% of the time the CPU clock is on 1190mhz, and never higher.
The second Benchmark is while running a Benchmark tool. CPU is used to 99% and always busy working on 1500mhz.

So yeah undervolting is more effective on higher voltages. But that's not the reason here.
While undervolting certainly does help, you just won't profit from it in everyday usage. Only when doing some really heavy usage. Most games probably won't even use that much CPU power.
14th February 2016, 06:22 AM |#150  
Member
Thanks Meter: 21
 
More
i see, well thank you for the replay and an awesome thread
12th March 2016, 01:53 AM |#151  
Senior Member
Thanks Meter: 22
 
More
good thread , thanks to the OP

by the way , I love the Z2 battery efficiency , and it also undervolts a lot (around -100mv) without stability issues

here's my voltages table :
frequency(mhz) : voltage(mV)
268 : 625
300 : 640
422 : 650
652 : 665
729 : 675
883 : 690
960 : 700
1036 : 710
1267 : 740
1497 : 770
1574 : 780
1728 : 800
1958 : 840
2150 : 885
2265 : 900

all test performed with stability test

Read More
Post Reply Subscribe to Thread

Guest Quick Reply (no urls or BBcode)
Message:
Previous Thread Next Thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes