Related
Are there any reports from anyone out there that actually killed their phone/fried their cpu from overclocking. I would search...but it's not working
I don't know about the CPUs used in cellphones, but how many burned out CPUs would you like? (I know someone who uses them as parts of his grunge art, and he actually buys them from computer repair shops for about a buck a piece to make sure he has enough.) It's not as common these days, but back when the "art" was in its infancy people burned them out all the time. And when vendors stole the AMD fans and substituted cheaper ones - back in P2 days, IIRC - there were more burned out AMD chips than working ones. (Normal clock on an undercooled chip is the same as overclocking a properly cooled chip.)
So yes, it can probably burn out the CPU unless you keep the overclocking to a minimum, which won't buy you anything noticeable in performance. I laugh when I see people posting that they're overclocking a 3.2GHz CPU to 3.3GHz. If that saves them 2 seconds a year they're lucky.
Well never heard of anybody burning there smartphones CPU, but there are a few app like SetCPU, where u can set profiles for when the temp of ur processor is to high to downclock it.
Also i have heard of people overclocking there phones to twice there normal power and never had any issues.
Theres no need to literally burn the cpu, but you can easily shorten the life of the phone
If you have control of the voltage, and can take it beyond the CPU's limits, it should be quite easy to burn out your CPU.
With that said though, most apps and kernels have safe limits in-place on the voltage and clocks you can set. For example, I can't pass 1.4Ghz with 1500mv (if that's the right term) on my Galaxy Tab. With that voltage anyway, anything above 1.4Ghz is unstable
espionage724 said:
If you have control of the voltage, and can take it beyond the CPU's limits, it should be quite easy to burn out your CPU.
With that said though, most apps and kernels have safe limits in-place on the voltage and clocks you can set. For example, I can't pass 1.4Ghz with 1500mv (if that's the right term) on my Galaxy Tab. With that voltage anyway, anything above 1.4Ghz is unstable
Click to expand...
Click to collapse
Very true. I know we are on the subject of phones, but most of today's computer motherboards have that 'safety' feature where if you overclock or the processor gets too hot, then it will shut down the computer. I am sure phones have the same similar feature.
If someone would teach me that would be great!
Sent from my SAMSUNG-SGH-I717 using xda app-developers app
Better battery life. But be careful with undervolting. It can cause phone to be unstable. Under clocking is more forgiving. If you are going to do either I would suggest only making small changes and then use phone for several hours to settle in. If all good then try next step lower and test until you find the limit of your phone. Then bump it back up to last stable setting.
There are no noticeable benefits.
jimmer411 said:
There are no noticeable benefits.
Click to expand...
Click to collapse
I agree, its mostly the placebo effect regarding lag, a variable effect regarding battery life, due to user apps, and downright dangerous regarding overheating.
But many people whine when its not available for a kernel. I don't have a very high opinion of its necessity.
Battery life in our phones is dominated by the screen and radio (cell and wifi) but I find that undervolting at least reduces the heat when I'm doing something intensive like SNES emulation. Lag in our phones is also primarily a software issue (I have other Android devices with the same hardware running at the same MHz and it's a smoother experience).
Not that my Note got hot, it's just less warm.
ZacksBuilds said:
If someone would teach me that would be great!
Sent from my SAMSUNG-SGH-I717 using xda app-developers app
Click to expand...
Click to collapse
There's no reason not to under volt.
UC=Under clock
UV=Under volt
OC=Over clock
OV=Over volt.
Newer generations of chips have tremendous tolerances. If you were to look at the S3, a Qualcomm built SoC that was a bit long-in-the-tooth by the time the Note was released, you'd see that in many cases, there were phones that featured the chip, but were clocked lower... I.E at 1,000 MHz.
Most chips are designed to run faster than they do, at least for short periods of time. Modern chips use temperature and load to designate the running frequency.
When Qualcomm releases a chip, they need to be sure that 99% of the units they produce, can run with a near-flawless level of stability. Because of this, they often run chips slower, and at a higher voltage than is actually needed. The result is, that Qualcomm or Samsung might design a chip rated for 2ghz, and release their first device with the chip running at 1.5ghz. This is common. Each piece of silicon is slightly different due to manufacturing differences. Chips are made on giant discs, with hundreds of chips on a single disc. The chips in the center are generally considered to be of a slightly higher quality, requiring less voltage, or capable of running at a higher frequency, sometimes both. Because of this, each device is capable of slightly more, or slightly less. The challenge is for Qualcomm to set the speed and power usage for every chip-- the challenge for us, is to see if we can run the chip at a higher frequency, while drawing less power....because power =heat, and power= battery life.
There are two different ways to approach customizing frequency and voltage to your needs.
One: If you could run your Note at 1,900mhz, instead of 1500mhz, and draw the same amount of power, that'd be considered a success case. Your device would be faster, without a hit in battery life.
Two: Or, if you could run the device at the same speed, and lower the voltage (saving battery life), this would also be considered a success.
Sometimes, but rarely in the mobile space, you'll get lucky and receive a prime piece of silicon, and you'll be able to over clock, WHILE undervolting...resulting in a faster phone that uses battery life.
Ranger was correct...clock speed changes on the order of +/-20 percent are hardly noticeable. However, power requirements grow exponentially at higher frequencies, and on a mobile device, they're noticeable. While a device might require 1.2 volts at 1.5 GHz, and 1.3 at 1.6ghz, it might also only require 1.05, or 1.0 volts at 1.4ghz. These numbers are theoretical, and shouldn't be used-they arent even close to correct. They're used merely to illustrate that to over clock, chips often require more power than over clocking is worth, in heat (and heat past a certain point will instantly fry your device, or will reduce the lifespan of the chip).
These are general ideas on over clocking. If you want to know something specific, ask away. When over clocking/underclocking, the governor used (a set of conditions that tells the CPU when to change from, say 300mhz (when the screen is off, and the device isn't being used), to 1.5ghz, when both cores are being used, and fully loaded.
If you want to save battery life, a combination of build.prop radio tweaks, modem tweaks, under clocking profiles, and using a custom governor can significantly increase standby time, and noticeably, but not too significantly increase usage time. Its important to know though, that the screen is the hungriest part of our device, and brightness is the the quickest setting to adjust to gain battery life.
The same is true for performance. Kernel, OC profiles, build.prop and launcher tweaks, GOU over clocking, etc. All in combination can have a noticeable effect, because performance is the net sum of dozens of moving parts.
Sent from my SGH-I717 using XDA Premium HD app
Jamesyboy said:
There's no reason not to under volt.
UC=Under clock
UV=Under volt
OC=Over clock
OV=Over volt.
Newer generations of chips have tremendous tolerances. If you were to look at the S3, a Qualcomm built SoC that was a bit long-in-the-tooth by the time the Note was released, you'd see that in many cases, there were phones that featured the chip, but were clocked lower... I.E at 1,000 MHz.
Most chips are designed to run faster than they do, at least for short periods of time. Modern chips use temperature and load to designate the running frequency.
When Qualcomm releases a chip, they need to be sure that 99% of the units they produce, can run with a near-flawless level of stability. Because of this, they often run chips slower, and at a higher voltage than is actually needed. The result is, that Qualcomm or Samsung might design a chip rated for 2ghz, and release their first device with the chip running at 1.5ghz. This is common. Each piece of silicon is slightly different due to manufacturing differences. Chips are made on giant discs, with hundreds of chips on a single disc. The chips in the center are generally considered to be of a slightly higher quality, requiring less voltage, or capable of running at a higher frequency, sometimes both. Because of this, each device is capable of slightly more, or slightly less. The challenge is for Qualcomm to set the speed and power usage for every chip-- the challenge for us, is to see if we can run the chip at a higher frequency, while drawing less power....because power =heat, and power= battery life.
There are two different ways to approach customizing frequency and voltage to your needs.
One: If you could run your Note at 1,900mhz, instead of 1500mhz, and draw the same amount of power, that'd be considered a success case. Your device would be faster, without a hit in battery life.
Two: Or, if you could run the device at the same speed, and lower the voltage (saving battery life), this would also be considered a success.
Sometimes, but rarely in the mobile space, you'll get lucky and receive a prime piece of silicon, and you'll be able to over clock, WHILE undervolting...resulting in a faster phone that uses battery life.
Ranger was correct...clock speed changes on the order of +/-20 percent are hardly noticeable. However, power requirements grow exponentially at higher frequencies, and on a mobile device, they're noticeable. While a device might require 1.2 volts at 1.5 GHz, and 1.3 at 1.6ghz, it might also only require 1.05, or 1.0 volts at 1.4ghz. These numbers are theoretical, and shouldn't be used-they arent even close to correct. They're used merely to illustrate that to over clock, chips often require more power than over clocking is worth, in heat (and heat past a certain point will instantly fry your device, or will reduce the lifespan of the chip).
These are general ideas on over clocking. If you want to know something specific, ask away. When over clocking/underclocking, the governor used (a set of conditions that tells the CPU when to change from, say 300mhz (when the screen is off, and the device isn't being used), to 1.5ghz, when both cores are being used, and fully loaded.
If you want to save battery life, a combination of build.prop radio tweaks, modem tweaks, under clocking profiles, and using a custom governor can significantly increase standby time, and noticeably, but not too significantly increase usage time. Its important to know though, that the screen is the hungriest part of our device, and brightness is the the quickest setting to adjust to gain battery life.
The same is true for performance. Kernel, OC profiles, build.prop and launcher tweaks, GOU over clocking, etc. All in combination can have a noticeable effect, because performance is the net sum of dozens of moving parts.
Sent from my SGH-I717 using XDA Premium HD app
Click to expand...
Click to collapse
So what I understood is this: instead of having let's say, your note running at 1.2ghz consuming 15% battery per hour you can have it at 1.5ghz consuming the same battery life?
Sent from my SAMSUNG-SGH-I717 using xda app-developers app
Hey, this is my first post here! **Edited for... uh... clarity and to add some information. Hopefully it's more readable now.
I found a way to control and manage the thermal throttling of the device.***
Modern smartphones create a lot of heat, and usually the casing of the device or the aluminum midframe chassis is used as a heatsink. Unlike a laptop, it does not have a cooling fan, so it must rely on the passive dissipation of heat through the casing of the device and the display panel to keep the CPU temperature down. On the Nexus 6P, there is decent thermal contact between the processor IC and the midframe. However, due to both the fact that the RAM is layered over top of the CPU and because the thermal contact is still not ideal, it is difficult to keep the CPU as cool as a computer implementation.
In order to reduce heat production and control the temperature of the device, the OEM implements thermal engine for Qualcomm MSM chipset in order to slow the frequency of the CPU and GPU cores down when the temperature is high.
Right now it is set to 50C throttling temperature, presumably so that the heat production does not cause the display panel to heat up to the touch, as at lower powers, the thermal mass of the aluminum frame plus passive dissipation will make case temperature increases not very noticeable. However, at 50C, the CPU will often throttle even at medium load, because the hardware does not make it easy to keep the temperature under that. This has been true for every phone I have worked with in the past and some are even worse at this.
Personally, I find that having a high (>45C) surface temperature is not a huge problem (You can decide that for yourself. if it is a problem for you, this won't help you. You can't change the power efficiency limit of the processor). In my use case, when I have a lot of applications open, especially Firefox in desktop mode, or 4-way Android M multitasking or something like that (or just typing to a lot of people in Facebook Messenger or something similar). I kinda use the thing like a computer so obviously it has extended periods of high CPU workload and the device starts to throttle back to something like 1344 or 960MHz or even lower.
Given this, what I want to do is change the CPU throttle temperature to a higher one. Generally, we don't need to worry about protecting the processor hardware because it has built in thermal shutdown/reset functions should something go wrong. It's at something like 110C, which sounds high, but as someone who uses a MacBook Pro, it is normal to see the CPU temperature that high! Intel generally throttles at a much higher temperature because they don't care about the actual temp of the heat sink, only that of the processor die.
In the past kernel I tried, there was an option to set up the thermal throttling temperature (God's Kernel). However, I switched to the AK kernel recently, due to its High Performance Audio feature. This kernel did not include support for this configuration.
I am running MH19Q Marshmallow stock.
I used File Explorer with root access to go to system/etc I think and there is a file called thermal_engine.conf or similar. If you edit it, you see there are a lot of values. Actually basically if you look around, there are a lot of temperature values in there seperated by the things they control. I would like to explain more, but I think it is better if you can open up your file and my file side by side and see for yourself what's different. The gist is that there's a table of values and a bunch of actions to take when they're hit, and of course, there are release temperatures, which are basically the lower hysterisis limit I think. The temperature values look like 44000 or 43000 by default, which means 44C and 33C (celcius)and I changed mine to 97000 (97 C)
Here you can find the content of the file. Type in the pastebin website, then put slash QYhi05rE.
I didnt keep my old file.... sorry about that... Perhaps someone can post it if they have it.
With stuff set to 97 C, the device heats up a lot more, obviously, but it's manageable. If you have something like Cinema 4K open for a long time, of course it will get to like 50C on the surface (That is quite unconfortable to put your hand on, but I'm okay with it). Hangouts video calling seems to be the worst and sometimes the battery will get higher than 50C and then stop charging. Given the design of the phone, by the time the battery gets too hot to be safe, the system will probably shutdown or restart, and you'll notice it LONG before anything becomes a problem.
Thanks for looking!
***Do this at your own risk, as with all root mods and tricks. Obviously this has the risk of breaking things or causing hardware to fail. High temperatures on BGA soldered chips have been observed to increase the failure rates, even in stuff like routers and TVs and other stuff that you don't generally think of as having thermal issues. My last phone (Note 5) kinda broke after a little while, although I'm not sure if me doing this caused it. (Appears to be display panel issue, but have not tested). All I know is that earlier that day I was outside filming on it and processing video, and that the area above the SoC got rather warm to the touch. Which should be read as "painfully hot" to most.
file removed? add disclaimer pls
LarryChendragon2099 said:
Hey, this is my first post here! **Edited for... uh... clarity and to add some information. Hopefully it's more readable now.
I found a way to control and manage the thermal throttling of the device.***
Modern smartphones create a lot of heat, and usually the casing of the device or the aluminum midframe chassis is used as a heatsink. Unlike a laptop, it does not have a cooling fan, so it must rely on the passive dissipation of heat through the casing of the device and the display panel to keep the CPU temperature down. On the Nexus 6P, there is decent thermal contact between the processor IC and the midframe. However, due to both the fact that the RAM is layered over top of the CPU and because the thermal contact is still not ideal, it is difficult to keep the CPU as cool as a computer implementation.
In order to reduce heat production and control the temperature of the device, the OEM implements thermal engine for Qualcomm MSM chipset in order to slow the frequency of the CPU and GPU cores down when the temperature is high.
Right now it is set to 50C throttling temperature, presumably so that the heat production does not cause the display panel to heat up to the touch, as at lower powers, the thermal mass of the aluminum frame plus passive dissipation will make case temperature increases not very noticeable. However, at 50C, the CPU will often throttle even at medium load, because the hardware does not make it easy to keep the temperature under that. This has been true for every phone I have worked with in the past and some are even worse at this.
Personally, I find that having a high (>45C) surface temperature is not a huge problem (You can decide that for yourself. if it is a problem for you, this won't help you. You can't change the power efficiency limit of the processor). In my use case, when I have a lot of applications open, especially Firefox in desktop mode, or 4-way Android M multitasking or something like that (or just typing to a lot of people in Facebook Messenger or something similar). I kinda use the thing like a computer so obviously it has extended periods of high CPU workload and the device starts to throttle back to something like 1344 or 960MHz or even lower.
Given this, what I want to do is change the CPU throttle temperature to a higher one. Generally, we don't need to worry about protecting the processor hardware because it has built in thermal shutdown/reset functions should something go wrong. It's at something like 110C, which sounds high, but as someone who uses a MacBook Pro, it is normal to see the CPU temperature that high! Intel generally throttles at a much higher temperature because they don't care about the actual temp of the heat sink, only that of the processor die.
In the past kernel I tried, there was an option to set up the thermal throttling temperature (God's Kernel). However, I switched to the AK kernel recently, due to its High Performance Audio feature. This kernel did not include support for this configuration.
I am running MH19Q Marshmallow stock.
I used File Explorer with root access to go to system/etc I think and there is a file called thermal_engine.conf or similar. If you edit it, you see there are a lot of values. Actually basically if you look around, there are a lot of temperature values in there seperated by the things they control. I would like to explain more, but I think it is better if you can open up your file and my file side by side and see for yourself what's different. The gist is that there's a table of values and a bunch of actions to take when they're hit, and of course, there are release temperatures, which are basically the lower hysterisis limit I think. The temperature values look like 44000 or 43000 by default, which means 44C and 33C (celcius)and I changed mine to 97000 (97 C)
Here you can find the content of the file. Type in the pastebin website, then put slash QYhi05rE.
I didnt keep my old file.... sorry about that... Perhaps someone can post it if they have it.
With stuff set to 97 C, the device heats up a lot more, obviously, but it's manageable. If you have something like Cinema 4K open for a long time, of course it will get to like 50C on the surface (That is quite unconfortable to put your hand on, but I'm okay with it). Hangouts video calling seems to be the worst and sometimes the battery will get higher than 50C and then stop charging. Given the design of the phone, by the time the battery gets too hot to be safe, the system will probably shutdown or restart, and you'll notice it LONG before anything becomes a problem.
Thanks for looking!
***Do this at your own risk, as with all root mods and tricks. Obviously this has the risk of breaking things or causing hardware to fail. High temperatures on BGA soldered chips have been observed to increase the failure rates, even in stuff like routers and TVs and other stuff that you don't generally think of as having thermal issues. My last phone (Note 5) kinda broke after a little while, although I'm not sure if me doing this caused it. (Appears to be display panel issue, but have not tested). All I know is that earlier that day I was outside filming on it and processing video, and that the area above the SoC got rather warm to the touch. Which should be read as "painfully hot" to most.
Click to expand...
Click to collapse
Not new.
This has been around and discussed for a while.
I have been running a modified thermal-engine.conf since day one.
Hello Axon 7 users, I just picked up one a couple of days ago. After finally figuring out the bootloader, bootstack and general stock experience I tested a little bit of gaming. I found that a basic game like Clash Royale heats the battery up to around 42°C already with low brightness and slow charging. A more intensive game like the new Knives Out runs only slightly hotter but it becomes apparent that CPU gets throttled soon after loading to 1036MHz across all cores causing lag.
It's disappointing so I tried to find how to modify the throttling. Using ZTE's Power Manager setting on performance or balanced doesn't seem to have a noticeable difference.I tried the only stock custom kernel AX7 but it's outdated on B32 and I find it randomly reboots regularly. The stock kernel itself allows some configuration, but the thermal settings in Kernel Adiutor don't reflect any charge.
A quick Google search brings up how LG V20 Snapdragon 820 users edit /system/etc/thermal-engine.conf to tweak the throttling levels. Their config is quite different but they mod big to 1824Mhz and let little scale itself.
I couldn't get thermal-engine.conf to use the thermal-engine-8996-perf.conf values by copying the values to it as it suggests inside. I tried renaming it with the -zte.conf ending as it suggests as well but that didn't work. After just renaming both the normal and perf conf files with a .bak ending, I've found better throttling performance. Big now throttles to 1632Mhz and little to 1324Mhz. As far as I can understand the files don't have charging rates inside, just GPU and CPU throttling.
However as expected the device heats up a few degrees more now. This now puts my battery up to 47°C in Knives Out under the same conditions. Charging is stopped at 45°C by the system so as previously mentioned it's unmodified.
I just wanted to check since I couldn't find it mentioned. Is everyone ok with gaming performance limited to 1036Mhz with the normal throttle? Also are my temperatures normal? I guess CPU doesn't seem that high reaching around 65°C, it's just that the battery has less than 20°C difference in intensive performance. I suppose it's a quirk of the heat pipe to battery as heatsink design. I just expected more from a metal unibody chassis and at least normal CPU gaming performance. I thought my Sony Z3 Compact design was bad for battery thermals, with the battery stacked behind the CPU board, sandwiched in insulating glass. But I didn't expect to see a phone to route a heatpipe directly to it's battery.
Anyway it is what it is. Follow this information if you want some better gaming performance at the cost of your battery cycle life. In my case I bought the Axon7 just as a separate media consumption device rather than a phone so I can live with the tradeoff. If battery gets bad enough before 2 years I'll consider using warranty at the loss of receiving their refurbished replacement. Manufacturer warranty's in fact cover batteries for 80% depletion.
I recommend the app DevCheck Pro for being able to monitor CPU, GPU, temperatures and other things overlayed. I think some others may do similar but they may not be updated for Big Little and are more instrusively overlayed.
Infy_AsiX said:
A quick Google search brings up how LG V20 Snapdragon 820 users edit /system/etc/thermal-engine.conf to tweak the throttling levels. Their config is quite different but they mod big to 1824Mhz and let little scale itself.
I couldn't get thermal-engine.conf to use the thermal-engine-8996-perf.conf values by copying the values to it as it suggests inside. I tried renaming it with the -zte.conf ending as it suggests as well but that didn't work. After just renaming both the normal and perf conf files with a .bak ending, I've found better throttling performance. Big now throttles to 1632Mhz and little to 1324Mhz. As far as I can understand the files don't have charging rates inside, just GPU and CPU throttling.
Click to expand...
Click to collapse
I read half of that to be honest, but just one thing: To make things harder, ZTE added added a write protection on the system. To disable it you have to use a computer and connect your phone with ADB, then issue "adb reboot disemmcwp" (like DISable EMMC Write Protection). Otherwise all the changes that you made get undone after a reboot, and obviously you'd have to reboot after modifying that file
On LOS you can use BeastMode (even if your phone isn't an A2017U) which for me is the best friggin kernel I've used in performance terms. There you can change thermal limits
Infy_AsiX said:
Hello Axon 7 users, I just picked up one a couple of days ago. After finally figuring out the bootloader, bootstack and general stock experience I tested a little bit of gaming. I found that a basic game like Clash Royale heats the battery up to around 42°C already with low brightness and slow charging. A more intensive game like the new Knives Out runs only slightly hotter but it becomes apparent that CPU gets throttled soon after loading to 1036MHz across all cores causing lag.
It's disappointing so I tried to find how to modify the throttling. Using ZTE's Power Manager setting on performance or balanced doesn't seem to have a noticeable difference.I tried the only stock custom kernel AX7 but it's outdated on B32 and I find it randomly reboots regularly. The stock kernel itself allows some configuration, but the thermal settings in Kernel Adiutor don't reflect any charge.
A quick Google search brings up how LG V20 Snapdragon 820 users edit /system/etc/thermal-engine.conf to tweak the throttling levels. Their config is quite different but they mod big to 1824Mhz and let little scale itself.
I couldn't get thermal-engine.conf to use the thermal-engine-8996-perf.conf values by copying the values to it as it suggests inside. I tried renaming it with the -zte.conf ending as it suggests as well but that didn't work. After just renaming both the normal and perf conf files with a .bak ending, I've found better throttling performance. Big now throttles to 1632Mhz and little to 1324Mhz. As far as I can understand the files don't have charging rates inside, just GPU and CPU throttling.
However as expected the device heats up a few degrees more now. This now puts my battery up to 47°C in Knives Out under the same conditions. Charging is stopped at 45°C by the system so as previously mentioned it's unmodified.
I just wanted to check since I couldn't find it mentioned. Is everyone ok with gaming performance limited to 1036Mhz with the normal throttle? Also are my temperatures normal? I guess CPU doesn't seem that high reaching around 65°C, it's just that the battery has less than 20°C difference in intensive performance. I suppose it's a quirk of the heat pipe to battery as heatsink design. I just expected more from a metal unibody chassis and at least normal CPU gaming performance. I thought my Sony Z3 Compact design was bad for battery thermals, with the battery stacked behind the CPU board, sandwiched in insulating glass. But I didn't expect to see a phone to route a heatpipe directly to it's battery.
Anyway it is what it is. Follow this information if you want some better gaming performance at the cost of your battery cycle life. In my case I bought the Axon7 just as a separate media consumption device rather than a phone so I can live with the tradeoff. If battery gets bad enough before 2 years I'll consider using warranty at the loss of receiving their refurbished replacement. Manufacturer warranty's in fact cover batteries for 80% depletion.
I recommend the app DevCheck Pro for being able to monitor CPU, GPU, temperatures and other things overlayed. I think some others may do similar but they may not be updated for Big Little and are more instrusively overlayed.
Click to expand...
Click to collapse
I have noticed the same performance many months ago.
I tried changing the thermal values with both ways through the conf file or a custom kernel but all implementations seem to be faulty as nothing changed.
In the end I gave up because I couldn't find a solution for this.
But I figured because my games clash of clans, ppsspp, gba emulators don't lag I din't care much.
If you find a solution let me/us know.
Or post the modded confs you're using as well if you can.
That's all from me.
I just renamed both the thermal-engine files with a .bak extension. I've also got ZTE's Power Manager frozen as the performance profiles there don't seem to do anything and I don't use it's other features. There's some kind of CPU GPU throttle still in place but it's much higher as previously mentioned,. After searching further I saw your discussion about /vendor/bin related throttle, maybe that's the fallback it's now on.
The device does get uncomfortably hot with a new demanding game at maximum settings. I wouldn't recommend doing this if you want to maintain your battery. However if you're interested I discovered the Ax7 allows defining a lower maximum battery voltage in another TL/DR post https://forum.xda-developers.com/showpost.php?p=74746734&postcount=1353. To explain simply, it's possible to limit the voltage low for health and safety while keeping the device almost primarily powered by mains. Effectively the battery is at an optimum low voltage, practically idle but very hot. A little complicated sure, but worth it. Getting a Daydream V1 tomorrow to play with, this stuff will help with heat and performance a lot. If anyone wants my long winded explanation, give me a shout.
The CPU temp does jump around higher than 70. I'm tending to think that current powerful mobile processors aren't efficient enough for the physical body constraints of phones. Let alone poorly designed ones. The 820 is meant to be an improvement over the 810, wouldn't believe it by the throttle required and performance lost. The 835 is efficient enough apparently. From experience though I have my doubts on reviews and benchmarks to reflect real usage stress.
edit: Oh and disable VDD restriction in your kernel setting if you've set it to auto enable. That seems to be a switch for the aggressive throttle still available after mod.
Sent from my ZTE Axon 7 using XDA Labs
Infy_AsiX said:
I just renamed both the thermal-engine files with a .bak extension. I've also got ZTE's Power Manager frozen as the performance profiles there don't seem to do anything and I don't use it's other features. There's some kind of CPU GPU throttle still in place but it's much higher as previously mentioned,. After searching further I saw your discussion about /vendor/bin related throttle, maybe that's the fallback it's now on.
The device does get uncomfortably hot with a new demanding game at maximum settings. I wouldn't recommend doing this if you want to maintain your battery. However if you're interested I discovered the Ax7 allows defining a lower maximum battery voltage in another TL/DR post https://forum.xda-developers.com/showpost.php?p=74746734&postcount=1353. To explain simply, it's possible to limit the voltage low for health and safety while keeping the device almost primarily powered by mains. Effectively the battery is at an optimum low voltage, practically idle but very hot. A little complicated sure, but worth it. Getting a Daydream V1 tomorrow to play with, this stuff will help with heat and performance a lot. If anyone wants my long winded explanation, give me a shout.
The CPU temp does jump around higher than 70. I'm tending to think that current powerful mobile processors aren't efficient enough for the physical body constraints of phones. Let alone poorly designed ones. The 820 is meant to be an improvement over the 810, wouldn't believe it by the throttle required and performance lost. The 835 is efficient enough apparently. From experience though I have my doubts on reviews and benchmarks to reflect real usage stress.
edit: Oh and disable VDD restriction in your kernel setting if you've set it to auto enable. That seems to be a switch for the aggressive throttle still available after mod.
Click to expand...
Click to collapse
That's weird... what are the ambient temps where you live? Here it's anything between 20 and 30 degrees and mine never gets that hot, and it barely throttles. Of course you shouldn't game while charging, that WILL throttle the phone.
I have a big old CPU heatsink without a fan, and when I charge the phone at night I just put it upon the heatsink. It keeps the battery around the ambient temp, which I guess helps with battery degradation.
A nice app for monitoring the CPU is Trepn profiler, you can program it to show you anything like frequencies and temps on 2 separate graphs for example
For you gamers out there
Finally I solved the thermal throttling problem in this lovely device.
BEWARE **YOUR PHONE WILL HEAT MORE**
first using your root explorer go to system/etc, you will find 3 files related to thermal throttling:
1. thermal-engine-8996-normal.conf
2. thermal-engine-8996-perf.conf
3. thermal-engine.conf
These files related to the configuration in the settings->power manager->power plan.
If you chose performance then your phone will use the perf.conf
If you chose Smart power-save your phone will use normal.conf
Edit:
After further investigation it seems that no matter what setting I choose the file that is active always the normal.conf.
perf.conf seems not affecting anything at all. Needs futher investigation.
That is the mechanic, now for the fun part:
Open the file you wish to edit using text editor.
The key is in these two sections
Code:
[SKIN_CPU_MONITOR]
#algo_type monitor
sampling 1000
sensor emmc_therm
thresholds 59000 62000
thresholds_clr 56000 60000
actions cluster0+cluster1 cluster0+cluster1
action_info 1400000+1700000 1500000+1824000
override 5000
[SKIN_GPU_MONITOR]
#algo_type monitor
sampling 1000
sensor emmc_therm
thresholds 41000 42000
thresholds_clr 39000 40000
actions gpu gpu
action_info 510000000 401800000
override 5000
As you can see the values above are a little bit different from yours, it's because I already change a bit.
The parts you should take consideration are thresholds, thresholds_clr, actions, and action_info.
Thresholds means in what temperature the device should throttle
Thresholds_clr means in what temperature the device will stop throttling
Actions mean which item to be throttled
Cluster0 = Little CPU
Cluster1 = Big CPU
GPU = GPU
so for example above
thresholds 59000
thresholds_clr 56000
actions cluster0+cluster1
action_info 1400000+1700000
Means it will start throttling at 59 celcius and stop throttling at56 celcius, the item to be throttled are little CPU which become 1400 mhz and big CPU which become 1700 mhz
You can add many parameters just by using spaces like example above.
Okay, hope it helps. If you confuse, just use my values above and unleash the 60fps beast. Muahahaha
Edit:
This is my current setting
Code:
[SKIN_CPU_MONITOR]
#algo_type monitor
sampling 1000
sensor tsens_tz_sensor11
thresholds 48000 50000
thresholds_clr 45000 49000
actions cluster0+cluster1 cluster0+cluster1
action_info 1500000+1824000 1400000+1700000
override 5000
Instead of emmc thermostat I use tsens_tz_sensor11 which is the temperature of the CPU core itself, I think it is better
Finally someone make magisk module for this (thanks nfsmw_gr), not yet tried it myself though
https://forum.xda-developers.com/showpost.php?p=76793058&postcount=14
I've already moved on to stock b01 oreo rom, now it is located in /system/vendor/etc/thermal-engine.conf
Here is my current config
Code:
[SKIN_CHG_LIMIT]
algo_type monitor
sampling 1000
sensor emmc_therm
thresholds 39000 41000
thresholds_clr 38000 40000
actions battery battery
action_info 1 2
[SKIN_CPU_MONITOR_NORMAL]
algo_type monitor
sampling 1000
sensor emmc_therm
thresholds 36000 39000 42000
thresholds_clr 35000 38000 41000
actions cluster0+cluster1 cluster0+cluster1 cluster0+cluster1
action_info 1500000+1824000 1400000+1700000 1300000+1438000
override 5000
[SKIN_GPU_MONITOR_NORMAL]
algo_type monitor
sampling 1000
sensor emmc_therm
thresholds 41000 42000
thresholds_clr 39000 40000
actions gpu gpu
action_info 510000000 401800000
override 5000
Thank you so much.
Is this valid for destroying the phone without voiding the warranty?
aLexzkter said:
Is this valid for destroying the phone without voiding the warranty?
Click to expand...
Click to collapse
Lol, I'm pretty sure there's a failsafe for temps
You'd probably want to also undervolt if you are removing thermal throttling to give yourself a little more headroom. Hellsgate kernel supports undervolting
Been doing similar for months and mentioned about it a few times in various places.
I don't think using the CPU temperature sensor is suitable for throttling as it peaks largely and fluctuates rapidly with corresponding load. The reason emmc_therm (quiet_therm in AOSP ROMs) is probably used for a suitable consistent reading. The related default temperature throttles need to be adjusted to reflect the changed sensor and it's nature. What will happen on your config is sudden throttling on CPU load that will unthrottle rapidly when load is decreased but the battery/device isn't likely any cooler.
Btw on AOSP I can't get any changes to stay after rebooting with disemmcwp system writable too. I'm guessing the kernel or something regenerates the defined values at startup. I found having the file modified through a Magisk module works however.
Sent from my Xperia Z3C using XDA Labs
I know this sounds super lazy but can this possibly be made into a magisk module for convenience for and for ease of enabling/disabling
dalebaxter01 said:
I know this sounds super lazy but can this possibly be made into a magisk module for convenience for and for ease of enabling/disabling
Click to expand...
Click to collapse
I would like this as well if it's not much hassle.
dalebaxter01 said:
I know this sounds super lazy but can this possibly be made into a magisk module for convenience for and for ease of enabling/disabling
Click to expand...
Click to collapse
put your changed file into this template https://forum.xda-developers.com/mi-5/how-to/how-to-disable-thermal-throttling-t3636574
Infy_AsiX said:
Been doing similar for months and mentioned about it a few times in various places.
I don't think using the CPU temperature sensor is suitable for throttling as it peaks largely and fluctuates rapidly with corresponding load. The reason emmc_therm (quiet_therm in AOSP ROMs) is probably used for a suitable consistent reading. The related default temperature throttles need to be adjusted to reflect the changed sensor and it's nature. What will happen on your config is sudden throttling on CPU load that will unthrottle rapidly when load is decreased but the battery/device isn't likely any cooler.
Btw on AOSP I can't get any changes to stay after rebooting with disemmcwp system writable too. I'm guessing the kernel or something regenerates the defined values at startup. I found having the file modified through a Magisk module works however.
Sent from my Xperia Z3C using XDA Labs
Click to expand...
Click to collapse
Yeah, this is also what I found. At first I just looked at which temp is the highest and use that sensor for the trigger. But then when I use cpu-z to observe the reading, I found that emmc_therm is the most consistent and not fluctuate drastically so much. Therefore I'm back using emmc_therm right now
Seeing this thread gives me flashback in the Android M days where a xposed module was needed to disable a system task killer that would kill my games if they used too much battery once I turn off screen _-_
Never again will I use my phone gaming while overheating message comes up.
It almost killed my phone, not turning on and charging. I had to disassemble the phone and remove the battery to make it charge again.
otaconremo said:
Never again will I use my phone gaming while overheating message comes up.
It almost killed my phone, not turning on and charging. I had to disassemble the phone and remove the battery to make it charge again.
Click to expand...
Click to collapse
You should always keep your eyes on the temp. My rule of thumb is under 55 degree is ok, above that is no no. If your games always reach above 55 then you should re-tweak your thermal config.
Use dev check pro to observe the temp, though it doesn't have widget for emmc_therm only the cpu therm, you can still see what's going on under the screen
Okay, so after reading through the comments I made a magisk module for myself to test, using the template from the Mi 5 thread.
I modified all 3 files to be sure it works (my device boots so yeah, try it and let me know).
Also made a copy of the modded files at /system/etc to cover all scenarios because in the ROM I run the files are located in /system/vendor/etc, not in /system/etc as in the OP .
It shouldn't affect anything I think.
Anyway without stalling you'll find the file attached at the bottom, as always you flash at your own risk, have a nice day!
(P.S. If this is ok I'll put it on my thread with my Jojoc mod.)
otaconremo said:
Never again will I use my phone gaming while overheating message comes up.
It almost killed my phone, not turning on and charging. I had to disassemble the phone and remove the battery to make it charge again.
Click to expand...
Click to collapse
No idea how you got that. I've never seen a overheating message on stock N or RR-O. I've gotten the device to overheat shutdown plenty of times after increasing the secondary CPU throttle (hasn't been described here) from 95degC to 105degC. Even on fan with enough load at maximum frequencies the CPU temp can spike to the presumedly 120degC Qualcomm shutdown (dropping one step on big cores is fine). This secondary throttle behaves differently by interrupting performance and unthrottling rapidly. Still boots fine after. I've yet to try if undervolting will allow that final big frequency step, on fan, increased to 105degC, no secondary throttle and no shutdowm.
I'd guess instead your battery voltage dropped too severely low being old degraded and too high a current draw, requiring a reset of the battery PCB or something.
iyancoolbgt said:
You should always keep your eyes on the temp. My rule of thumb is under 55 degree is ok, above that is no no. If your games always reach above 55 then you should re-tweak your thermal config.
Use dev check pro to observe the temp, though it doesn't have widget for emmc_therm only the cpu therm, you can still see what's going on under the screen
Click to expand...
Click to collapse
55 isn't high at all for a CPU. You mean floating monitor overlay? It can select two custom temperatures to show on the dashboard and float monitor. The only real concern is battery temp. Other components are designed and rated to handle high temperatures. FWIW I've been running around 100degC CPU, keeping battery <50degC on fan with battery voltage limited to 3.85V for several months playing various high end games at maximum performance without notable ill effect. Well except that constant heat on the battery even at the safe low voltage will cause degradation, unavoidable cost for higher performance on a phone.
nfsmw_gr said:
Okay, so after reading through the comments I made a magisk module for myself to test, using the template from the Mi 5 thread.
I modified all 3 files to be sure it works (my device boots so yeah, try it and let me know).
Also made a copy of the modded files at /system/etc to cover all scenarios because in the ROM I run the files are located in /system/vendor/etc, not in /system/etc as in the OP .
It shouldn't affect anything I think.
Anyway without stalling you'll find the file attached at the bottom, as always you flash at your own risk, have a nice day!
(P.S. If this is ok I'll put it on my thread with my Jojoc mod.)
Click to expand...
Click to collapse
That location sounds like how it was when I tried AEX-O. Don't know if anything's changed there but no files could affect ROM thermal throttling on it then. The files there are different too, maybe generic Sd820 config or something, whereas RR-O uses the same sensors, values and config as stock N. I did also try the stock config modded to AEX but still no go. The only file with effect where it works is the -normal.conf.
Infy_AsiX said:
No idea how you got that. I've never seen a overheating message on stock N or RR-O. I've gotten the device to overheat shutdown plenty of times after increasing the secondary CPU throttle (hasn't been described here) from 95degC to 105degC. Even on fan with enough load at maximum frequencies the CPU temp can spike to the presumedly 120degC Qualcomm shutdown (dropping one step on big cores is fine). This secondary throttle behaves differently by interrupting performance and unthrottling rapidly. Still boots fine after. I've yet to try if undervolting will allow that final big frequency step, on fan, increased to 105degC, no secondary throttle and no shutdowm.
I'd guess instead your battery voltage dropped too severely low being old degraded and too high a current draw, requiring a reset of the battery PCB or something.
55 isn't high at all for a CPU. You mean floating monitor overlay? It can select two custom temperatures to show on the dashboard and float monitor. The only real concern is battery temp. Other components are designed and rated to handle high temperatures. FWIW I've been running around 100degC CPU, keeping battery <50degC on fan with battery voltage limited to 3.85V for several months playing various high end games at maximum performance without notable ill effect. Well except that constant heat on the battery even at the safe low voltage will cause degradation, unavoidable cost for higher performance on a phone.
That location sounds like how it was when I tried AEX-O. Don't know if anything's changed there but no files could affect ROM thermal throttling on it then. The files there are different too, maybe generic Sd820 config or something, whereas RR-O uses the same sensors, values and config as stock N. I did also try the stock config modded to AEX but still no go. The only file with effect where it works is the -normal.conf.
Click to expand...
Click to collapse
Mind sharing your own config?
Infy_AsiX said:
No idea how you got that. I've never seen a overheating message on stock N or RR-O. I've gotten the device to overheat shutdown plenty of times after increasing the secondary CPU throttle (hasn't been described here) from 95degC to 105degC. Even on fan with enough load at maximum frequencies the CPU temp can spike to the presumedly 120degC Qualcomm shutdown (dropping one step on big cores is fine). This secondary throttle behaves differently by interrupting performance and unthrottling rapidly. Still boots fine after. I've yet to try if undervolting will allow that final big frequency step, on fan, increased to 105degC, no secondary throttle and no shutdowm.
I'd guess instead your battery voltage dropped too severely low being old degraded and too high a current draw, requiring a reset of the battery PCB or something.
55 isn't high at all for a CPU. You mean floating monitor overlay? It can select two custom temperatures to show on the dashboard and float monitor. The only real concern is battery temp. Other components are designed and rated to handle high temperatures. FWIW I've been running around 100degC CPU, keeping battery <50degC on fan with battery voltage limited to 3.85V for several months playing various high end games at maximum performance without notable ill effect. Well except that constant heat on the battery even at the safe low voltage will cause degradation, unavoidable cost for higher performance on a phone.
That location sounds like how it was when I tried AEX-O. Don't know if anything's changed there but no files could affect ROM thermal throttling on it then. The files there are different too, maybe generic Sd820 config or something, whereas RR-O uses the same sensors, values and config as stock N. I did also try the stock config modded to AEX but still no go. The only file with effect where it works is the -normal.conf.
Click to expand...
Click to collapse
I believe my battery is too degraded already. That only happens when my batt is below 20% and I was still gaming. That's why I'm very careful already when I'm getting lowbat. Btw, I'm using stock chinese mf5 version. The overheating message is from the Mi-Assistant app (Power Manager).
nfsmw_gr said:
Mind sharing your own config?
Click to expand...
Click to collapse
I did awhile back but I guess it's a little scattered across posts. I was about to edit in another paragraph with more info but I'll put it here now instead. I'll consider cleaning up the new stuff to share, performance profiles is just a varied collection of performance levels.
I've setup various CPU GPU performance profiles to switch between using Tasker. From testing, the highest I'd go without a fan is 1600/1800 big and 1300 small. With a constant load at these frequencies the battery will still hit up to around 50degC, matching the much higher perf allowed with a fan.
Here's a tidbit to get y'all thinking about environmental impact on heat. Stock Daydream sets a Sustainable Performance mode that uses all cores at 1200MHz. In the Daydream Headset V1, the back of the phone actually has an air gap behind it because of raised nubs inside the headset. Yet with all this the battery can rather easily heat to over 55degC after a few dozen minutes. The headset v2 has a built in heatsink, with minor improvement apparently. Also on an off-topic note, that Sustained Performance mode gets stuck to on after using Daydream on stock N. Things no one notices unless you actually monitor
---------- Post added at 10:12 PM ---------- Previous post was at 10:09 PM ----------
otaconremo said:
I believe my battery is too degraded already. That only happens when my batt is below 20% and I was still gaming. That's why I'm very careful already when I'm getting lowbat. Btw, I'm using stock chinese mf5 version. The overheating message is from the Mi-Assistant app (Power Manager).
Click to expand...
Click to collapse
Yeah sounds like my guess about an undervolted degraded battery. Oh I had that frozen so never got that alert. I'd assume it occurs when battery is 55degC or so, that's when Daydream alerts too for example.
Infy_AsiX said:
No idea how you got that. I've never seen a overheating message on stock N or RR-O. I've gotten the device to overheat shutdown plenty of times after increasing the secondary CPU throttle (hasn't been described here) from 95degC to 105degC. Even on fan with enough load at maximum frequencies the CPU temp can spike to the presumedly 120degC Qualcomm shutdown (dropping one step on big cores is fine). This secondary throttle behaves differently by interrupting performance and unthrottling rapidly. Still boots fine after. I've yet to try if undervolting will allow that final big frequency step, on fan, increased to 105degC, no secondary throttle and no shutdowm.
I'd guess instead your battery voltage dropped too severely low being old degraded and too high a current draw, requiring a reset of the battery PCB or something.
55 isn't high at all for a CPU. You mean floating monitor overlay? It can select two custom temperatures to show on the dashboard and float monitor. The only real concern is battery temp. Other components are designed and rated to handle high temperatures. FWIW I've been running around 100degC CPU, keeping battery <50degC on fan with battery voltage limited to 3.85V for several months playing various high end games at maximum performance without notable ill effect. Well except that constant heat on the battery even at the safe low voltage will cause degradation, unavoidable cost for higher performance on a phone.
That location sounds like how it was when I tried AEX-O. Don't know if anything's changed there but no files could affect ROM thermal throttling on it then. The files there are different too, maybe generic Sd820 config or something, whereas RR-O uses the same sensors, values and config as stock N. I did also try the stock config modded to AEX but still no go. The only file with effect where it works is the -normal.conf.
Click to expand...
Click to collapse
Maybe I'm just too cautious But still, too much heat makes my hands really not comfortable.
I know we can customize which temperature to show, but no emmc_therm unfortunately. Reading your comment make me thinking, if the real concern is the battery temp, then why not using it as the trigger for the thermal config? It is listed as 'battery' in cpu-z list of thermal sensors. It means that we can use it right?
nfsmw_gr said:
Okay, so after reading through the comments I made a magisk module for myself to test, using the template from the Mi 5 thread.
I modified all 3 files to be sure it works (my device boots so yeah, try it and let me know).
Also made a copy of the modded files at /system/etc to cover all scenarios because in the ROM I run the files are located in /system/vendor/etc, not in /system/etc as in the OP .
It shouldn't affect anything I think.
Anyway without stalling you'll find the file attached at the bottom, as always you flash at your own risk, have a nice day!
(P.S. If this is ok I'll put it on my thread with my Jojoc mod.)
Click to expand...
Click to collapse
Thanks, that's so cool. I'll put it on the first post