Disabling thermal throttling on Note 4 Exynos - Samsung Gear VR

I was able to completely disable thermal throttling on N4 Snapdragon (including Oculus driver that throttles independently of kernel), it will work at full performance in VR until it blacks out from insufficient voltage. But now it's half dead because of EMMC fail (not caused by this mod - It will happen to every single N910F board after ~1.5 years because Samsung decided to install cheap memory).
It still boots if I cool it down with an ice pack and works fine after that but it won't last forever. I don't know exactly how I was able to completely disable thermal throttling - I was unable to reproduce the effect from fresh install, even after repeating all steps, so every time the system gets corrupted due to bad EMMC, I flash my lucky TWRP backup with no throttling. The backup is a Dr. Ketan ROM M12 with AEL kernel 10.8 I modified to work with ported N7 ROMs.
Now I want to make my N4 Exynos ignore temperatures and battery level in VR the same way.
Does anyone know how to achieve that? Simply adjusting thermal values in kernel is not enough, I need a way to disable throttling in Oculus Driver.
edit: I found some info on how it works, could be useful if anyone else is working on it https://developer.oculus.com/documentation/mobilesdk/latest/concepts/mobile-power-overview/
I'm fully aware of what I'm doing, please don't lecture me about possible dangers.

Related

OnePlus 3 [Encrypted vs Unencrypted] [Benchmark Results & Discussion]

Hey XDA.
So I'm guessing I'm not the only one wondering if the OnePlus 3 benefits or not from coming with storage encryption by default as per Android Marshmallow OEM requirements. So I put a little experiment together and decided to share my results with the XDA OP3 community.
Disclaimer
None of the results here should be taken as an absolute, as always these tests are flawed and should be taken with a grain of salt, the whole idea of this was to see if disabling encryption would boost "performance" on the device.
Testing Methodology
To attempt to reduce differences that could affect results between both testing scenarios (Encryption enabled and disabled) I followed a set of rules perfoming these tests:
Both scenarios had the phone with the same amount of data that is to say, same apps installed and files on sdcard as to replicate results on an equal and real environemnt (not freshly restored device with no data or apps)
Tests were run back to back for each set of tests. after the first set (Boot Time) there was a 3 minute period for the phone to cool down before perfoming the second set (Antutu) etc.
First all tests were run with encryption enabled after disabling encryption and restoring all apps and data a period of time (5-6min) was taken to ensure device was cool again before tests for the second scenario were started.
Boot Time was done by starting timer as soon as power button was held and holding until OnePlus Logo and Name appeared on screen, timer was stopped as soon as Lockscreen rendered.
Results
Oneplus 3
Encrypted------------------------------------------- Un-Encrypted
Boot Time
0:30:82-------------------------------------------0:30:32
0:34:44-------------------------------------------0:28:98
0:30:78-------------------------------------------0:29:54
0:34:79-------------------------------------------0:28:32
0:34:28-------------------------------------------0:28:15
Median 1,2,3,4,5
M 33:02-------------------------------------------M 29:06 +12%
Boot Time saw a 12% benefit from Unencrypted storage which I expected but in my opinion isn't as big as a difference when compared to other devices with and without encryption before the Snapdragon 820
Antutu
stuck at 70%------------------------------------ 143698
143297------------------------------------------- 143213
141531-------------------------------------------- 142985
141849------------------------------------------- 142895
141501------------------------------------------- 142502
Median 2,3,4,5
M 142045-------------------------------------------M 142899 +0.60%
The first test performed on the encrypted scenario got stuck at 70% and as to not skew results the first scores for both scenarios were ignored when calculating median, here we can see a 0.60% increase in performance for the unencrypted scenario which in my opinion is too small of a difference to draw a correlation between encrypted and unencrypted and worth mentionting is that anyway I believe no significant differences would be tested and found here between both scenarios because of the way encryption works.
Geekbench
2409-----------------------------------------2396
5620-----------------------------------------5597
2433-----------------------------------------2406
5591-----------------------------------------5628
2428------------------------------------------2392
5587----------------------------------------- 5578
2439-----------------------------------------2390
5643-----------------------------------------5581
2431------------------------------------------2391
5569-----------------------------------------5544
Median 1,2,3,4,5
M 2428-----------------------------------------M 2395 -1.35%
M 5602--------------------------------------------M 5586 -0.28%
Here we can observe a decrease in performance of 1.35% in single core results and 0.28% in multi-core results which again in my opinion does not amount to anything noticeable in day to day usage and I believe is not the result of encrypted or unencrypted storage differences
Click to expand...
Click to collapse
Conclusion
Based on these results and this personal experience only, I conclude that for me the only benefit of unencrypting my storage was a shorter boot time (~4sec) and performance wise there is little to no difference at all, and while the security benefits of Android encryption are questionable to say the least. some employers/companies require this to be enabled at all times, and since in these results the differences are so miniscule, there would be at a glance no reason to not stay encrypted if you already are. Although a more in depth testing methodology and cases where storage read/write speeds were predominant would be ideal to maybe see significant differences between performance of encrypted and unencrypted storage, I presume this would not affect a regular user enough to warrant the reformating of all user data if the storage is already encrypted, which will probably be the common state of Android phones that are coming out now and in the future.
Click to expand...
Click to collapse
Obviously this will not replicate the same experience anyone can have so I encourage anyone to reply to this thread with their personal experience on the matter, what is your opinion on encryption? and if you find this post helpful somehow remember to hit the thanks button on my post or any other user that replies helpfully to you. Thanks.
Android's Full-Disk Encryption (FDE) Can Be Cracked on Qualcomm-Based Devices
http://news.softpedia.com/news/andr...racked-on-qualcomm-based-devices-505900.shtml
****
"Although Beniamini is working with both Qualcomm as well as Google, the core of the issue might not be completely fixable and might even require new hardware changes to fix. "
http://thehackernews.com/2016/07/hacking-android-encryption.html
******
I don't know what to think.
Why encrypt when everyone and their mother can crack it?
Seems like our device will never have 'Apple-grade' encryption.
gruntyoldbag said:
Android's Full-Disk Encryption (FDE) Can Be Cracked on Qualcomm-Based Devices
http://news.softpedia.com/news/andr...racked-on-qualcomm-based-devices-505900.shtml
****
I don't know what to think.
Why encrypt when everyone and their mother can crack it?
Seems like our device will never have 'Apple-grade' encryption.
Click to expand...
Click to collapse
"According to Duo Security, 57% of all Android devices, from the company's internal data set, are vulnerable to this attack, the rest having Google's May Android security update installed, which patched CVE-2016-2431, or just don't use a Qualcomm chip."
so if you have the latest OOS which has the June 1st patch, your safe. For now. hackers are always one step ahead, but the chances of your or my device being hacked are slim to none. Look at stagefright, not one single incident.
Nice that they put the script on GitHub tho... And "Apple-grade encryption" means absolutely squat. How do you think iCloud images leak onto the internet?
Read the admentment to my post above.
Also iCloud images leaked because of weak passwords / guessing security questions on the web.
Certainly not because of a weak OS/Chipset.
gruntyoldbag said:
Read the admentment to my post above.
Also iCloud images leaked because of weak passwords / guessing security questions on the web.
Certainly not because of a weak OS/Chipset.
Click to expand...
Click to collapse
I believe that all systems are vulnerable to attackers at some level, even the ones considered the "most" secure (paypal, banks etc.) can and have been hacked.
Apple devices are no more or no less secure than android, (5.0+) both systems have their weaknesses and can be brute force attacked. A lot of security experts are suggesting the Nexus devices to be used if security is a concern as they at least get monthly updates. Nice thing about that is AOSP gets updated around the same time, so custom roms that are kept up to date with the latest merges reap the same benefits as a Nexus device.
FYI: I'm not trying to sound preachy/rude/short or anything, I had a heated argument this morning with my companies IT department as they still want us to use iPhone 5S/BBZ10's instead of new Nexus/Samsung devices.. (our plan is up for renewal)
Well if I understand correctly it requires bruteforcing the password/pin.... I suggest to encrypt but also use a strong password. There's no 100% safe security anyway. It depends of your 'adversary' too. For most users it's not the NSA. So even if the protection is not perfect a strong password should be good enough. You can use the fingerprint reader too (and switch off the phone in some situations, like crossing borders...).
It's a bit concerning that some responding to this OP are basically making light of the very, very serious vulnerabilities mentioned in the above links. Obviously Apple, at least as of the writing of the above articles, seems to have better system for FDE. Especially on their newer chipsets. I really hope that Qualcomm and Google figure out some sort of work around for this for our 820, but also the other QC SoCs that are currently supported.
No, my adversary is not the NSA or other intel agencies(hopefully). I would like a (highly?) reliable method of disk encryption that actually works and isn't compromised in detail on a blog post along with a whole bunch of scripts and code and the developer of hashcat posting in the comments section with goal of collaborating on way to further optimize the attack and integrate into the HC code base.
Call me paranoid but I'll never be any company's, corporation's, organization's or developer's fanboy... IMO I AM THE PAYING CUSTOMER. DELIVER OR **** OFF.
I love Android and OOS is a very true interpretation but as far as security goes this NEEDS to be fixed ASAP in one way or another. Too bad we can't get a copperOS ROM. Even if we had that for the OP3 it appears that the issues are largely the result of QC's implementation of the "TrustedZone".
How to disable encryption?
what would be more interesting are actual i/o speeds before and after encryption not crappy antutu and geekbench, im not gonna get horny of benchmarks from antutu like alot of people do, but read and write speeds are different, i might get a semi
how can i remove this Encryption from recovery????
It does not sound negligibile at all. Why didnt you test with large app starts and app installation. I, for instance, accidentally encrypted my op3t and i noticed by seeing that a game was being installed for an unusually long time.
Thanks. I'm not going to bother with trying to stay unencrypted from now on after seeing this. I always seem to mess the process up and end up encrypted, lol. Like my current ROM. I know the other day I looked and was not encrypted. Now I am. I have no idea what I did but I certainly didn't do it on purpose. I don't care about boot or install times since I don't do that much anyway.
Guys go for unencrypted gives best performance + f2fs for data and cahe both.

Looking for a successor or upgrade for my Samsung Galaxy Note SM-P900

Hello World,
I have been working (and sometimes labouring) with my trusty work horse tablet SM P900 for almost 18 months now. And despite numerous disappointments (USB OTG w/o charging?!), it has been a mostly satisfying ride.
But since more than a few weeks now, I am starting to notice a slow down in memory intensive applications. It is almost as if the interal storage is acting up whilst reaching its EOL.
Actions that are not CPU intensive at all seem to be hanging, until all of a sudden they break loose and everything seems to be working normal again - until next time. And this happens more and more frequently. I tried to "trim" myself out of this mess, but by now I can't see any real difference here no more.
I tried in vain to find a replacement board for the internal storage, but alas replacement parts for the P900 seem to be very, very hard to obtain in the first place.
If anyone here has some useful suggestion, then I all ears (BTW: I have been using a slimed down custom ROM from day one and I am very picky about what apps I allow to run in the background. So its absolutely no SW issue).
Now I am also willing to entertain the idea of a new purchase, but try as I may I could not find a worthy successor to the P900. Samsung itself seems to have given all but up on this line and all the competition doesn't life up to my expectations.
These are:
1) It *must* be running on USB 5V power - anything else would be a deal breaker for me.
2) It should be running on an ARM CPU, but I could do with an Intel as well.
In any case, it is ARM based, then there has to be some custom Android ROM available for it (I hate stock bloated stock firmware).
It it is Intel, then it needs to have Linux support - somewhere out there!
3) The Display should be HD and somewhat comparable with the P900 specs.
4) The unit must be tablet capable, best with notes like stylus support. So now fixes keyboard and the like.
5) For ARM/Android based RAM must be 3GB minimum and internal storage 64GB min.
6) There must be an option to insert an extra micro SDXC card (deal breaker).
If anyone knows of a manufacturer who can match my desires, I am very much open to suggestions right now.

Battery overheating tips

Soo I have a faulty battery on my Z3c and I do not have the time or money to replace it right now,and I would love to have a phone which wouldn't overheat and thermal throttle after watching a YouTube video and opening Snapchat.
As well because the battery overheated the back glass's adhesive has weakened and the glass unsticked so I use the phone with the battery exposed which actually did help a bit.
I do get better results when using "conservative" governor and at least trying to set the clock to 1500MHz(it maxes at 1500MHz for some time and then starts peaking at 2400MHz again). I also tried undervolting,didn't notice a difference unfortunately.
So now,what can I do to cool down the battery,because it is starting to get very very annoying. Maybe putting some aluminum plate over the battery(maybe a dumb suggestion,don't really know) or something like that? Thanks to anyone who helps in advance.
Mpdecision becoming re-enabled causes frequency's not to stick. I found using init.d to set kernel CPU settings and having Kernel Adiutor apply as normal again prevents issues.
See the last photo of my battery mod gallery for a thermal mod https://imgur.com/a/8Ceha. It still heats up, though taking longer to and a little less. You could also apply new CPU thermal paste, it wears out with age. Of two devices I've opened up one had no paste.
A metal back will dissipate heat better but will probably affect wireless signal. An interesting observation I've noticed is with the battery separated outside of the device the CPU tends to run a little hotter. As poor design as having a battery attached to the opposite side of a CPU acting as a heatsink is, it still does dissipate heat over more surface than without. More so I think with the battery removed there's not enough pressure against the back of the board to push the thermal contact flush with the front display body. With the back panel removed there's not enough pressure. With the air gap separation mod the extra ~1mm lets an external case create some pressure, getting better results than with an original glass back and hard plastic back case on top.
I never really ran my phone case-less so can't report differences there. However it should be noted that glass is an insulator and isn't good for dissipating heat.
Infy_AsiX said:
Mpdecision becoming re-enabled causes frequency's not to stick. I found using init.d to set kernel CPU settings and having Kernel Adiutor apply as normal again prevents issues.
See the last photo of my battery mod gallery for a thermal mod https://imgur.com/a/8Ceha. It still heats up, though taking longer to and a little less. You could also apply new CPU thermal paste, it wears out with age. Of two devices I've opened up one had no paste.
A metal back will dissipate heat better but will probably affect wireless signal. An interesting observation I've noticed is with the battery separated outside of the device the CPU tends to run a little hotter. As poor design as having a battery attached to the opposite side of a CPU acting as a heatsink is, it still does dissipate heat over more surface than without. More so I think with the battery removed there's not enough pressure against the back of the board to push the thermal contact flush with the front display body. With the back panel removed there's not enough pressure. With the air gap separation mod the extra ~1mm lets an external case create some pressure, getting better results than with an original glass back and hard plastic back case on top.
I never really ran my phone case-less so can't report differences there. However it should be noted that glass is an insulator and isn't good for dissipating heat.
Click to expand...
Click to collapse
Soo first the Mpdesicion thing,I have installed a new kernel which removes Mpdesicion and puts Intellithermal instead,I don't know if that affects anything with clock changes but we'll see. I am gonna switch to Oreo soon so we'll see how it is there.
Second,I will take the phone apart definitely when I get some time to do so. And for the third thing,no I am not really into modding cases and the insides of the phone and similar so I will be probably avoiding that.
Thanks for the help though.
Dinkec27 said:
Soo first the Mpdesicion thing,I have installed a new kernel which removes Mpdesicion and puts Intellithermal instead,I don't know if that affects anything with clock changes but we'll see. I am gonna switch to Oreo soon so we'll see how it is there.
Second,I will take the phone apart definitely when I get some time to do so. And for the third thing,no I am not really into modding cases and the insides of the phone and similar so I will be probably avoiding that.
Thanks for the help though.
Click to expand...
Click to collapse
Intellithermal is for thermal throttling. A hotplugging replacement for Mpdecision is something else.
So nothing is covering the battery? Be careful not to puncture it. A case doesn't require modding but I think some thickness is required to create pressure. In other words I think the back panel without a case is better than no back panel with just a case and no pressure modification.
Infy_AsiX said:
Intellithermal is for thermal throttling. A hotplugging replacement for Mpdecision is something else.
So nothing is covering the battery? Be careful not to puncture it. A case doesn't require modding but I think some thickness is required to create pressure. In other words I think the back panel without a case is better than no back panel with just a case and no pressure modification.
Click to expand...
Click to collapse
Okay I will try something today or tomorrow.
As for the Mpdecision,I turned it off and my clock still peaks at 2500MHz. You mentioned something with init.d.
Could you tell me what to do?
Dinkec27 said:
Okay I will try something today or tomorrow.
As for the Mpdecision,I turned it off and my clock still peaks at 2500MHz. You mentioned something with init.d.
Could you tell me what to do?
Click to expand...
Click to collapse
Your ROM needs to have init.d support. If it does there should be a folder in /system/etc/init.d/. This used to be easier to add on MM stock. Since N AOSP it seems to be disabled. I've got it working via a workaround on Magisk, it's not straightforward.
Maybe try enabling shell script to apply settings in Kernel Adiutor settings first. If that fails maybe search and ask to see how anyone else gets edited frequencies to stick in your kernel thread. If you still can't get a fix i'll recommend my ideas.
Infy_AsiX said:
Your ROM needs to have init.d support. If it does there should be a folder in /system/etc/init.d/. This used to be easier to add on MM stock. Since N AOSP it seems to be disabled. I've got it working via a workaround on Magisk, it's not straightforward.
Maybe try enabling shell script to apply settings in Kernel Adiutor settings first. If that fails maybe search and ask to see how anyone else gets edited frequencies to stick in your kernel thread. If you still can't get a fix i'll recommend my ideas.
Click to expand...
Click to collapse
So I did most of what you said,so first with Oreo it is actually much better with battery and heat management,but how do you now fix the volume down button when you tried to make the buttons go to place for 1 hour?
Dinkec27 said:
So I did most of what you said,so first with Oreo it is actually much better with battery and heat management,but how do you now fix the volume down button when you tried to make the buttons go to place for 1 hour?
Click to expand...
Click to collapse
So what have you done? Useful to know what's worked for you. Or do you mean Oreo is cooler by itself?
Do you mean your volume button is stuck? It shouldn't be. Check inside if it's aligned and properly placed. If you had removed the board try reinstalling it again taking care with button alignment.
Sent from my Z3 Compact using XDA Labs
Infy_AsiX said:
So what have you done? Useful to know what's worked for you. Or do you mean Oreo is cooler by itself?
Do you mean your volume button is stuck? It shouldn't be. Check inside if it's aligned and properly placed. If you had removed the board try reinstalling it again taking care with button alignment.
Click to expand...
Click to collapse
Nah button alignment is completely fine it just isn't registiring a key press,maybe I have damaged the flex cable or the board.
And what I have done is apply thermal paste,leave 1mm between the motherboard and the battery so it has some air to breathe and Oreo seems to be cooler itself.
You must disable Thermal Core Control for your settings to be applied, can do it with Kernel Adiutor
Well I managed to get a new battery and with new settings that I set on Oreo there is almost no heat sooo yeah,this thread can be closed.

CPU temp always 10.8-11 degree

Hello,
I dont have an idea why my CPU temp shows always something arround 10.8-11 degree. Is the sensor defect? I already made a factory reset and flashed actual firmware complete with odin. Any ideas?
It's not an temperature sensor but an headphone jack ADC. It don't know why they (Samsung kernel devs) marked it as thermal sensor.
BTW I made a new recovery that fixes all these issues: wrong date, cpu temperature sensor, encryption support, brightness settings. Will post it soon when finish theming =) and polishing

Question Display unreadable under strong sun

Hello.
After a few months with my Exynos S21 Ultra, totally happy with it, yesterday I went to the beach for the first time to discover a terrible issue: if the phone gets warm, under the sun, the display dims down becoming unreadable.
I immediately thought about a problem in my device, trying to restart it several times, putting it in safe mode, checking every possible menu or submenu.
Nothing.
I moved to a bar, inside, and after a little while, everything went back to normal.
I tried to Google the issue and discover this is something "NORMAL" for Exynos phones (found a thread about the S20 Ultra):
if the phone heats, the screen enters a sort of "safe mode" dimming itself down to an unreadable point.
This means you can barely take pics because you see nothing in sunlight.
This is something new for me because nothing similar ever happened on my previous Note 10.
Anybody?
Best
maxito71 said:
Hello.
After a few months with my Exynos S21 Ultra, totally happy with it, yesterday I went to the beach for the first time to discover a terrible issue: if the phone gets warm, under the sun, the display dims down becoming unreadable.
I immediately thought about a problem in my device, trying to restart it several times, putting it in safe mode, checking every possible menu or submenu.
Nothing.
I moved to a bar, inside, and after a little while, everything went back to normal.
I tried to Google the issue and discover this is something "NORMAL" for Exynos phones (found a thread about the S20 Ultra):
if the phone heats, the screen enters a sort of "safe mode" dimming itself down to an unreadable point.
This means you can barely take pics because you see nothing in sunlight.
This is something new for me because nothing similar ever happened on my previous Note 10.
Anybody?
Best
Click to expand...
Click to collapse
Not sure what anyone can do about this. This is not exclusive to the Exynos 2100, happens with the SD888 too (probably more often) and not just on the S21U, even the OP9Pro has similar thermal issues.
When overheated (either due to intense processor activity like Camera use, or external factors like ambient heat and sunlight) the phone will throttle down the CPU, close energy intensive processes like camera and dim down the screen among other things to minimize thermal damage.
It's not a pleasant experience, but it is what it is...
enigmaamit said:
Not sure what anyone can do about this. This is not exclusive to the Exynos 2100, happens with the SD888 too (probably more often) and not just on the S21U, even the OP9Pro has similar thermal issues.
When overheated (either due to intense processor activity like Camera use, or external factors like ambient heat and sunlight) the phone will throttle down the CPU, close energy intensive processes like camera and dim down the screen among other things to minimize thermal damage.
It's not a pleasant experience, but it is what it is...
Click to expand...
Click to collapse
Does this happen with the iPhone?
nixnixnixnix4 said:
Does this happen with the iPhone?
Click to expand...
Click to collapse
I don't know.
iPhone can mean many different models, just like "android phone" can mean many different devices.
Not all Android phones have the above issue. This seems to be an inherent issue with the current architecture of both the E2100 and the SD888. They run hot, especially on the S21U and the OP9Pro. Once in a while we get chips with less than desirable thermal personality, just like the SD810 had in the past. If it's going to overheat, it will sacrifice performance and features to facilitate cooling down.
That's all I was trying to say in my post. I know you're not a new member here, so I'm presuming your question was more rhetorical than literal. If something I said ticked you off, it wasn't intentional - I was just pointing out the way things are with this generation of processors. And I don't think it's user rectifiable without crippling device performance.
If you have a solution to modify the E2100/SD888 thermal behaviour, please let us know, I'm sure a lot of grateful users would like a phone that doesn't heat up under heavy load.

Categories

Resources