So I decided to tear my device open and see if I could install a copper shim on the CPU to make better contact with the metal frame. I took some pictures as well and I'll have a link to that below. I didn't have the time to go into a lot of testing but I did have some interesting results.
For starters I was surprised to see a cut out on the RF shielding for the CPU. The CPU actually already had some sort of silicone pad about .5 mm in thickness that contacted the frame. I was really surprised to see this since it's not common to see that on other phones with qualcomm chipsets. This probably explains why the phone gets so warm even with relatively low CPU temperatures sense the heat is being transferred to the frame already. I noticed this as well when I added a copper shim to my Galaxy Nexus. CPU temps dropped but the frame got warmer, after all the heat has to go somewhere.
Although it already had a thermal pad I decided to try to add a .5mm copper shim with diamond thermal paste since they're known to conduct heat better than thermal pads. You can see in the link below pictures of the teardown and copper shim install.
https://www.dropbox.com/sh/lw20gdye88welct/AAA0aH26jZbb41zSoJ1fjpjwa?dl=0
As far as results I just ran some benchmarks and compared them to results from before the copper shim. I don't have as much as I'd like to have so perhaps some of you can run some of these test on your device to see what results you get with no copper shim.
Antutu 26C no Shim
49,133
Antutu 26C with copper shim. I also let the it cool down for 10 seconds between each run.
1str run 54,604
2nd run 49270
3rd run 48769
4th run 46165
5th run 42739
6th run 43145
Geekbench 3
No copper shim
28 celsius soc temperature.
Single score 1137
Multi score 3724
With copper shim
28 celsius soc temperature.
Single Score 1200
Multi score 4085
2nd run
Single 1176
Multi 4057
3rd run
Single 1136
Multi 3247
4th run
Single 732
Multi 3060
5th Run
Single 698
Multi 2551
6th run
Single 702
Multi 2664
GFXbench 3.1
Manhattan ES 3.1
No copper Shim
Start temp 31C
End Temp 36.5C
GPU 100mhz drop after 50 seconds
GFXbench 3.1
Manhattan ES 3.1
With copper Shim
Start temp 31C
End Temp 35C
No GPU speed drop through the 60 second test.
So it did help but eventually it still thermal throttles quite a bit.
Thanks for sharing this. Your results are interesting.
@probaina Speaking out of experience from my days as CPU/GPU desktop overclocker, I recommend using less thermal paste. You should only really barely see a thin layer of paste on the die, especially when you are not able to apply any kind of pressure within a mobile device.
goldlocke said:
@probaina Speaking out of experience from my days as CPU/GPU desktop overclocker, I recommend using less thermal paste. You should only really barely see a thin layer of paste on the die, especially when you are not able to apply any kind of pressure within a mobile device.
Click to expand...
Click to collapse
I agree 100% less is more when it comes to thermal compound. The thinner the layer the better it will transfer the heat.
Sent from my LG-H950 using XDA Premium HD app
Intub8 said:
I agree 100% less is more when it comes to thermal compound. The thinner the layer the better it will transfer the heat.
Sent from my LG-H950 using XDA Premium HD app
Click to expand...
Click to collapse
I agree as well. I was just testing how much of a gap there was on that picture. It seems like .5mm wasn't exactly making enough contact for a thin layer of thermal paste. I actually tested two 5mm shims and then the thermal paste layer was a lot thinner seemed to delay thermal throttling a just bit longer as well. I'm just waiting on a 1mm shim to finally exchange it with. I'm now playing with the phone's thermal configuration settings. I'm trying to experiment with different settings with thermal hotplug and throttling of individual cores to find out what's a good compromise between heat and performance.
@probaina how long is the side of the shim 10mm ? 15mm?
I finally bought a 12x12mm and that fit perfectly. 1mm thickness worked better for me as well. 1mm of thickness needed a lot less thermal paste and just fills the gap better.
probaina said:
I finally bought a 12x12mm and that fit perfectly. 1mm thickness worked better for me as well. 1mm of thickness needed a lot less thermal paste and just fills the gap better.
Click to expand...
Click to collapse
Any difference in temperature?
Actually nothing that's very significant. I'll be posting new results soon.
probaina said:
I finally bought a 12x12mm and that fit perfectly. 1mm thickness worked better for me as well. 1mm of thickness needed a lot less thermal paste and just fills the gap better.
Click to expand...
Click to collapse
Where do you bought it, maybe a link please?
I got it from here http://pages.ebay.com/link/?nav=item.view&id=271865367802&alt=web
Although I got another new LG G Flex 2 from AT&T to compare and benchmark side by side. For some reason this new device runs almost as good as with a thermal shim. I'll make another post with new results.
I got another new G Flex 2 to replace my current yellow screen one. So now I can do some testing with and without a thermal Shim comparison with another device side by side. Also on the plus side the screen on my new device has no yellow at all and colors just look better. I'm not sure if this newer device is better built or what but it seems to have better thermal cooling than what I recall my first G Flex 2 having. Maybe one of you want to test this on a non AT&T model and let me know your results. I'll probably take it apart and see if anything is different later. For now here are the results. I did two tests running Geekbench 3 with no rest in between. Both devices where running the default Thermal configuration file. I measured device temperature, scores, and also measured frequency with Trepn Profiler. I ran the first test with both devices starting the benchmark with a temperature of 26C.
1st Benchmark
W/shim No Shim
1308 1214
3878 3741
1233 1179
3690 3705
1204 1172
3510 3658
969 1106
3342 3303
681 894
3004 3173
700 722
2812 2927
708 692
2884 2883
705 698
2658 2674
Final temperature:
41C 40C
2nd Benchmark
1348 1228
3820 3882
1295 1207
3775 3661
1222 1150
3457 3440
1109 928
3439 3246
881 667
3149 2820
670 702
2814 2639
706 703
2570 2512
Final Temperature:
41C 41C
These results seem very similar but when I compared the frequency graph I can see the big cores stayed at a higher frequency for longer than the device with no copper shim. The graph only shows big core activity to keep things simple. I've attached the graphs if you want to compare.
I have to admit that this new device I got from AT&T does seem to have better thermal capability than my original one did. Just running Antutu it scored 55,500 which my original device never even came close to before adding the copper shim. So either this newer phone has a better thermal pad or it just runs cooler. I'll be opening this new device today.
So after taking this new device apart I can see that's there's nothing different about it. However I did notice the thermal pad had evidence of making better contact with the metal frame. It was obvious on the pad and the frame. Even then it looks like a bit over half of the pad was really touching the frame. On my first G Flex 2 most of the thermal pad was smooth with no pressure indentations so that probably explains why it throttled easier. So if you're lucky to get a device with a thermal pad that's making good frame contact then it'll perform close to as good as the copper shim. It's just unfortunate that LG didn't use a slightly thicker thermal pad to guarantee good thermal contact with all devices.
I saw a video that showed a teardown of the LG G4 and it clearly shows extensive modifications to the SOC's thermal transfer pad. The layout is very similar to the G Flex 2 but there's a strange gold mesh on top of the SOC. It's also protruding some as if maybe there's a pad between the SOC and the gold mesh. On the frame where the thermal pad contacts the frame, it now has grey pad that sticks out some. This would definitely be a huge improvement to the really bad thermal pad contact our devices have. I've attached pictures that I've found of the G4’s thermal pads.
Taking a look at the LG G4 thermal configuration it shows the little cluster dropping only to a minimum of 48C vs 42C on ours. The G4 big cluster throttles by default at 53C but only drops to a minimum of 51C vs ours of 48C max and a really low 38C minimum. Even their big cluster hot plug settings run hotter for longer. They have a similar maximum temperature at which they hot plug as our 5 and 6 core but their cores quickly come back on at 54C and 56C. In our devices the cores stay off until the temperature goes down to 45C.
I read an article on Arstechnica that says that it looks like the Snapdragon 808 in the LG G4 runs cooler because it won’t thermal throttle as much. To me it looks like it runs hotter looking at the configuration file. It’s most likely not thermal throttling as much because of the combination of more reliable thermal transfer to the frame, lower clock speed and thermal configuration that allows the SOC to not thermal throttle as much.
I just found this interesting and though I’d share. However it really makes me wonder if the Snapdragon 810 would perform better if it came with the same thermal pad parts, as well as configured with a thermal configuration that allows the SOC to run hotter as in the Snapdragon 808.
great job man! you really did a great job for Flex 2.
spiderx_mm said:
great job man! you really did a great job for Flex 2.
Click to expand...
Click to collapse
Thanks a lot. Although I feel like the thermal configuration file tweaking helped out more then this mod. I'm just glad that LG allows the thermal settings to be changed. I just love doing this stuff so if it can help someone out there it's even better.
probaina said:
I saw a video that showed a teardown of the LG G4 and it clearly shows extensive modifications to the SOC's thermal transfer pad. The layout is very similar to the G Flex 2 but there's a strange gold mesh on top of the SOC. It's also protruding some as if maybe there's a pad between the SOC and the gold mesh. On the frame where the thermal pad contacts the frame, it now has grey pad that sticks out some. This would definitely be a huge improvement to the really bad thermal pad contact our devices have. I've attached pictures that I've found of the G4’s thermal pads.
Taking a look at the LG G4 thermal configuration it shows the little cluster dropping only to a minimum of 48C vs 42C on ours. The G4 big cluster throttles by default at 53C but only drops to a minimum of 51C vs ours of 48C max and a really low 38C minimum. Even their big cluster hot plug settings run hotter for longer. They have a similar maximum temperature at which they hot plug as our 5 and 6 core but their cores quickly come back on at 54C and 56C. In our devices the cores stay off until the temperature goes down to 45C.
I read an article on Arstechnica that says that it looks like the Snapdragon 808 in the LG G4 runs cooler because it won’t thermal throttle as much. To me it looks like it runs hotter looking at the configuration file. It’s most likely not thermal throttling as much because of the combination of more reliable thermal transfer to the frame, lower clock speed and thermal configuration that allows the SOC to not thermal throttle as much.
I just found this interesting and though I’d share. However it really makes me wonder if the Snapdragon 810 would perform better if it came with the same thermal pad parts, as well as configured with a thermal configuration that allows the SOC to run hotter as in the Snapdragon 808.
Click to expand...
Click to collapse
Thanks for your effort ,did you test the lg g4 thermal to gflex2?I made the changes you mentioned above i still havent spotted much difference.i dont know if qualcomm can provide us more info
kutulu32 said:
Thanks for your effort ,did you test the lg g4 thermal to gflex2?I made the changes you mentioned above i still havent spotted much difference.i dont know if qualcomm can provide us more info
Click to expand...
Click to collapse
I've been wanting to test the G4 thermal settings as well. Although they really don't look that aggressive so I didn't think it would make a big difference. It may be that running at 1.8 Ghz max helps as well. I also have no clue how much cooler the Snapdragon 808 in the G4 really runs. I can't find any tests that show how much cooler the G4 runs compared to the G Flex 2 that I can compare with. The one test I did find was tested on a Geekbench beta app that's not available to the public. I did however order an LG G4 for my wife and when it gets here this week I'll be doing a side by side runs with it with my LG G Flex 2.
At least we know that changing the thermal settings on the Flex 2 can significantly lower thermal throttling and boost performance by a lot without making the device much warmer. Surprisingly it runs better than I ever expected with the thermal mods. I also can't find any phone that won't thermal throttle after a while. Even when Arstechnica tested thermal throttling on the S6 it thermal throttled down to 1.4ghz after 330 seconds of benchmarking. There seems to be a lot of criticism towards the Snapdragon 810 but I kind of blame more of this on the settings than on the chip. Under normal use I find that the Flex 2 feels cooler to the touch than my S4 or G2 did. I guess when the G4 gets here I'll know if it's really that much better or not.
I agree that the software/settings is that causes the most issues not the cpu. Lg tries to save us from overheating lol
http://developer.lge.com/resource/mobile/RetrieveBootloader.dev?categoryTypeCode=ANRS
LG can no longer guarantee the full functionality of your device after you unlock the bootloader. Unlocking your device may cause unexpected side effects that may include but are not limited to the following:
- Your device may stop working.
- Certain features and functionalities may be disabled.
- Your device may become unsafe to the point of causing you harm.
- Your device becomes physically damaged due to overheating.
No other comment..... I hope that experienced devs will build custom kernels for our device after the forthcoming bootloader unlock.
Related
Way high. Anyone having issues is not making the right yes no questions about LGs location tracing service and the stupid McAfee bloatware install. I just owned pretty much everything with my scores. No need root.
No need disable cores.
SnapDragon 810 is fine. Users are stupid as usual.
Antutu is scoring over what iphone 6 should too... Over 47000 something. You have to remember unless you root and tweak this phone it is a dialed down snapdragon 810. You have to bring it back to full speed to get full speed benchmarks that would match and beat a galaxy s6. For whatever reason LG did that, it definitely does make it slower out of the box than the s6 in benchmarks. But after you de-crap your unrooted version. It runs plenty fast and not laggy.
optimatic said:
Antutu is scoring over what iphone 6 should too... Over 47000 something. You have to remember unless you root and tweak this phone it is a dialed down snapdragon 810. You have to bring it back to full speed to get full speed benchmarks that would match and beat a galaxy s6. For whatever reason LG did that, it definitely does make it slower out of the box than the s6 in benchmarks. But after you de-crap your unrooted version. It runs plenty fast and not laggy.
Click to expand...
Click to collapse
The SD810 should be getting scores above 50,000 on Antutu and as for mopping the floors with everything the GS6 scores around 69,000. The 810 would score well over what my current benchmarks are, which is around 52,000 if there weren't heating issues.
If I let the cpu cool down to 20C I can get an antutu score of 58,000. I'm going to see if I can add a copper shim and thermal compound. It really made a big difference when I did that with my old galaxy nexus.
DIY manual PLZ!!!
probaina said:
If I let the cpu cool down to 20C I can get an antutu score of 58,000. I'm going to see if I can add a copper shim and thermal compound. It really made a big difference when I did that with my old galaxy nexus.
Click to expand...
Click to collapse
The problem is when you run it a second or third time you get a score that's comparable to galaxy s5 scores lol
guys, please.
You own Snapdragon 810, 64-bit TOP Qualcomm processor. It's perfect with his on-paper specs, so let's try to make him perfect in real.
The only thing matters is the UI lagging. Do you really meet your friends and say "Yo, dude, my phone took 58k points in Antutu?" No, you say "dude, it's fast" or not fast.
It's G Flex 2 - it's alredy curve, stylish and sexy. It points attention to itself. So is there a huge difference between 58k or 48k points in Antutu? Does this really matter, when you like your phone?
If you really need that points - go and buy that awkward, terribly looking SGS6.
I'm on this board because I like to tweak with my phone. I enjoy pushing hardware and seeing what its capable of. I didn't really buy my LG Flex 2 to show off but I'm pretty sure most of us are here because we already like our phones.
probaina said:
I'm on this board because I like to tweak with my phone. I enjoy pushing hardware and seeing what its capable of. I didn't really buy my LG Flex 2 to show off but I'm pretty sure most of us are here because we already like our phones.
Click to expand...
Click to collapse
This. I love doing this stuff. Especially when a phone has such physical appeal as the Flex 2. We may have out of the box issues, but that's why we're here. To iron out those issues!
Sent from my LG-H950 using XDA Free mobile app
dark.wizard said:
guys, please.
You own Snapdragon 810, 64-bit TOP Qualcomm processor. It's perfect with his on-paper specs, so let's try to make him perfect in real.
The only thing matters is the UI lagging. Do you really meet your friends and say "Yo, dude, my phone took 58k points in Antutu?" No, you say "dude, it's fast" or not fast.
It's G Flex 2 - it's alredy curve, stylish and sexy. It points attention to itself. So is there a huge difference between 58k or 48k points in Antutu? Does this really matter, when you like your phone?
If you really need that points - go and buy that awkward, terribly looking SGS6.
Click to expand...
Click to collapse
I love my phone... Until it gets hot, starts going unbareably slow and I have to sit there holding the power button for 30 seconds for the phone to reboot because rebooting normally would take about 5 to 10 minutes due to throttling. The phone is gorgeous but runs far from it. "We" are looking for ways to push it so it doesn't have these problem. Going with the mentality that "oh it's qualcomms best so it's amazing" mentality is ignorant. No offense but it is :/... Until issue are fixed with it the 808 and 805 run better because of speed and reliability (less lag) I am also a member of XDA to push my device and make it better. So no I'm not happy with this half baked chip until we come up with a way to fully bring out it's potential and share it with the community.
That is really odd because somehow I'm not experiencing lag. No matter how hard I push my phone it never seems to lag. I have the AT&T version of this phone. Could it be possible that somehow the AT&T version is different? From what I've been reading I haven't heard any one with the AT&T version complain about lag. Even when I try to push my phone hard I never see it go past 37C. If I run benchmarks for a long time like Antutu four times in a row the highest I've seen is 47C and even then my phone doesn't lag at all. So I'm wondering what's different on the phones with users that experiencing that much lag.
They may have throttled down the A57s to 1.5ghz. They've been doing that with them to try to fix the heat issues
In my AT&T model my A57 cores run at 2ghz but at around 40c it disables two cores and drops them to 1.55ghz.
The att version comes with 3 gigs of ram instead of 2
Yes but that shouldn't cause such a dramatic performance difference.
Of course it can. Most people talk about they can factory reset their device and it runs smooth as butter. Sounds like the phone starts to chug when it is loaded down with user data and apps with background processes.
I guess it can depending on how much stuff is running in the background. I know on my Galaxy S4 with 2GB of ram it runs smooth with cm11 android 5.1 no matter how many things are open. Although it also doesn't have the bloatware that comes with an OEM device.
probaina said:
That is really odd because somehow I'm not experiencing lag. No matter how hard I push my phone it never seems to lag. I have the AT&T version of this phone. Could it be possible that somehow the AT&T version is different? From what I've been reading I haven't heard any one with the AT&T version complain about lag. Even when I try to push my phone hard I never see it go past 37C. If I run benchmarks for a long time like Antutu four times in a row the highest I've seen is 47C and even then my phone doesn't lag at all. So I'm wondering what's different on the phones with users that experiencing that much lag.
Click to expand...
Click to collapse
I have EU version and I have no lag as well with v10e. Phone runs around 40C when gaming or browsing. There is some delay when multitasking, but no significant lag.
Some phones are great to take camping because if you play Asphalt 8 long enough, the back warms up to the ideal temperature that can bake bread. Rate this thread to express the extent to which the Huawei P10 Plus stays cool under extended heavy use. A higher rating indicates that even when playing strenuous games for long periods of time, the phone doesn't get uncomfortably warm.
Then, drop a comment if you have anything to add!
due to metal body when its start warm up, the whole phone is hot.
I know they wanted this because it cooling down on the bigger surface.
I just cant imagine how much heat inside the phone.
it never shut down itself and felt no issue with the heat, just surprised.
The first few months of using the phone and regularly playing mostly of power intensive and heavy games (Real Racing, GT Racing 2, Asphalt, Online MOBAS) has given me excellent outputs in terms of graphics and lag-less experience. The heat however, can be expected to be felt much more due to its metallic build and there are times that the screen heats up as well. But in my own experience on the P10 plus, its a tolerable heat than the other phones i have handled before (Huawei P8 Lite, Iphone 5s, Samsung S5) and records using cpuz shows heat signatures not exceeding 40 degrees, high but can be tolerable enough. During peak heat signatures while on gaming, it does not give you a decrease in performance nor drops of frame-rates that can decrease your gaming experience (but while playing Real Racing, I cant lock in to 60fps always and it drops to minimum of 35 fps in some occasions). Bought some plastic leather flip cover and the heat cant be felt at all.
EDIT:
Doing intensive camera shoots and video records, I did not felt any significant high heat signatures BUT using FB and FB messenger long enough gives you high heat signatures which for me is weird.
jsedayao said:
The first few months of using the phone and regularly playing mostly of power intensive and heavy games (Real Racing, GT Racing 2, Asphalt, Online MOBAS) has given me excellent outputs in terms of graphics and lag-less experience. The heat however, can be expected to be felt much more due to its metallic build and there are times that the screen heats up as well. But in my own experience on the P10 plus, its a tolerable heat than the other phones i have handled before (Huawei P8 Lite, Iphone 5s, Samsung S5) and records using cpuz shows heat signatures not exceeding 40 degrees, high but can be tolerable enough. During peak heat signatures while on gaming, it does not give you a decrease in performance nor drops of frame-rates that can decrease your gaming experience (but while playing Real Racing, I cant lock in to 60fps always and it drops to minimum of 35 fps in some occasions). Bought some plastic leather flip cover and the heat cant be felt at all.
EDIT:
Doing intensive camera shoots and video records, I did not felt any significant high heat signatures BUT using FB and FB messenger long enough gives you high heat signatures which for me is weird.
Click to expand...
Click to collapse
If your phone heats up by using FB and FB Messenger then it's a firmware bug or something since I use these a lot and mine does not even get warm at all. it's ice cold still. But if I play games for a while it gets room warm temperature not really hot or so as well. Could be a bug with cores, normally most cores should be IDLE when not needed but on yours it might be so when you launch FB and FB Messenger it thinks you're running a heavy game so it wakes up the idle cores so all cores are active. While on mine most are 0% = IDLE and only goes on when needed, such as when playing a game or so.
Expanding battery issue
Since about a week I noticed that the back of my P10 Plus was slightly bent outwards. Is it possible that the battery could have expanded and bent the metal or is it something else. I'm a little bit worried my phone will explode
Edit: It also always gets warm even if I'm just browsing the web or reading an article. I called my provider where I got the phone from and they said I might have a defect battery
Edit 2: I compared it to my OnePlus One. When doing the same thing (boot up, load 3 websites, 1 min of instagram, and 1 min of Spotify) my P10 Plus' battery was 36.0°C while my OPO's battery was 28.6°C. Both without a case P10 was at 64% and opo at 88%
After a few months of using P10 Plus, I am still bitter about the quality of this phone. While I was on vacation using Google Maps (Mobile data, GPS, screen on) in 33+ degrees C under the burning sun, the phone was very uncomfortable to hold. Now that I am back home where it's more cool, it still gets too hot, but not to the point where I am almost burning my hand. An annoying problem, probably as a result of the heat, is that graphics performance in games gets worse after a short while. I also took the screen protector off in order to apply "nano liquid", but the 37 USD worth of "nano liquid" kits simply didn't work on this screen. I could use my previous phone (Galaxy Note II) for 5 years, but this P10 Plus (my most expensive phone) is something I wanted to replace a few days after purchase.
Can anyone make a kernel with overclock. It's probably the most reasonable device for OC because of large surface area, aluminum and proper cooling interface. But development is zero. I was running all cores in performance governor and it doesn't affect temperature or battery in noticeable way. I am sure it would be beneficial to OC. Is there an easy way to do it myself? After seeing that tab s7 is noticeably worse everywhere and overall: weight wise, battery, life screen, speakers and microphone it's obvious that there is no need to pay for downgrade. We can topple up our cpu close to 865 level which looks as mostly verison with higher clock.
That would be awesome, I guess eventually someone may develop something.
As with the latest commitment of Samsung to support 3 years of SW updates this machine suddenly became a lot more future proof than I expected so together with an optimized and oc'ed kernel that would be awesome.
Great stuff.
Sora94 said:
As with the latest commitment of Samsung to support 3 years of SW updates this machine suddenly
Great stuff.
Click to expand...
Click to collapse
oh really? that's great
and yeah i wonder why no one develop custom kernel for tab s6, and note 10 too with the same cpu, but there are kernels with others like one plus etc
Since turning it on for the first time I often notice overheating in the upper area under the cameras. It does this very often when it is empty or when I use heavy apps like Call of Duty mobile .. has anyone noticed something similar? It seems absurd to me that such a phone will overheat like this
Sorry to hear that your phone is overheating.
Except when installing the phone the first time, i did not have any problems with it heating up. I dont use have apps so can't help you with that ( most heavy app i use is prob Clash of clans haha)
Unfortunately, that is the 888. Only real solution is to decrease the settings on "heavy" games and take breaks.
I have the same issue. My sister and cousin have the phone also but neither is experiencing it. I don't play games or watch videos on the phone. I only use it for calls, messaging, and online shopping so there is no reason for it to happen. I called to get a replacement. Hopefully the new device will not have the same issue.
Oh perfect! So if I replace I resolve?
Galaxy Z Flip3 5G, Phones Support | Samsung Care US
Samsung Care is here to help you with your Galaxy Z Flip3 5G. Get started, find helpful content and resources, and do more with your Samsung product.
www.samsung.com
Contact Us | Official Samsung Support US |
Samsung Care is here to help. You can connect with our Care Pros in a variety of ways. Choose the one that works for you.
www.samsung.com
I don't know how anybody would expect this very compact phone to not get hot with intensive tasks. It shouldn't get hot with light usage, though. Also try using Samsung browser instead of chrome if it gets hot when browsing.
I noticed mine getting awful hot the other day while using Android Auto on my car stereo with it in my pocket in my shirt, got so hot I had to take my phone out my pocket and put it in my cup holder. I noticed it getting hot also while watching Youtube videos at the house also, it wasn't in my pocket against my body though so it wasnt really that big a deal! LOL!
twistedumbrella said:
Unfortunately, that is the 888. Only real solution is to decrease the settings on "heavy" games and take breaks.
Click to expand...
Click to collapse
This is correct. Let's just be happy Samsung didn't put the SD888+ in the phone. The SD888 already had thermal problems. Can only imagine the SD888 being hotter from being overclocked.
Its not overheating, its getting hot. Android OS will shut down and give a warning when overheating. If your not using a case you will notice a lot more than a previous phone with a case installed. Its a CPU and GPU that creates heat. Its normal to get hot during high intensive cpu/gpu use and or when charging. It also tends to get hot when setting up for 1st time and transferring data from an old device.
JayRolla said:
Its not overheating, its getting hot. Android OS will shut down and give a warning when overheating. If your not using a case you will notice a lot more than a previous phone with a case installed. Its a CPU and GPU that creates heat. Its normal to get hot during high intensive cpu/gpu use and or when charging. It also tends to get hot when setting up for 1st time and transferring data from an old device.
Click to expand...
Click to collapse
I think you may have missed the point trying to correct the wording. The 888 gets significantly hotter than even the previous 865+ performing the same tasks. The issue is that what used to be just a bit warm is now hot and one warm day will easily push that into overheating. The performance increase may not justify the difference, which is where a lot of people will have an issue.
twistedumbrella said:
I think you may have missed the point trying to correct the wording. The 888 gets significantly hotter than even the previous 865+ performing the same tasks. The issue is that what used to be just a bit warm is now hot and one warm day will easily push that into overheating. The performance increase may not justify the difference, which is where a lot of people will have an issue.
Click to expand...
Click to collapse
Ive been using mine in 90* weather gaming, videos, etc... and have had the same warm/hot feeling I got with my s10+. I have yet to have overheating. I didnt see anyone mention the OS shutting down due to regular use and overheating but I could be mistaken. Being that I have been overclocking PC's, phones for 20 years the heat I am getting from my Flip 3 seems pretty normal.
JayRolla said:
Ive been using mine in 90* weather gaming, videos, etc... and have had the same warm/hot feeling I got with my s10+. I have yet to have overheating. I didnt see anyone mention the OS shutting down due to regular use and overheating but I could be mistaken. Being that I have been overclocking PC's, phones for 20 years the heat I am getting from my Flip 3 seems pretty normal.
Click to expand...
Click to collapse
I understand the point you are trying to make, but Android has not existed for 20 years. Anything before the G1 is irrelevant. That said, you also have to discard anything you used that wasn't Snapdragon, since Qualcomm and Mali (or others) are simply not the same thing.
Now that you are looking at a much less impressive list of past devices, you have to consider that you aren't actually overclocking and the 888 is currently capped. You also have to consider that the 888 does perform much more efficiently, so it may even seem cooler than an S10+ when doing the same basic tasks. After all, the 855 was known for getting hot and is now two generations back in terms of performance. When you notice the heat is when you run things the S10+ would struggle to handle.
The problem I was actually pointing out was comparing the 865+ and 888. The 865+ clocks higher than the 888, but the 888 runs significantly hotter. Running the same games on an 865+ and an 888 will yield higher temperature from the 888, but only marginal performance improvement. The question is what happens when you try to run the 888 at its full capabilities?
twistedumbrella said:
I understand the point you are trying to make, but Android has not existed for 20 years. Anything before the G1 is irrelevant. That said, you also have to discard anything you used that wasn't Snapdragon, since Qualcomm and Mali (or others) are simply not the same thing.
Now that you are looking at a much less impressive list of past devices, you have to consider that you aren't actually overclocking and the 888 is currently capped. You also have to consider that the 888 does perform much more efficiently, so it may even seem cooler than an S10+ when doing the same basic tasks. After all, the 855 was known for getting hot and is now two generations back in terms of performance. When you notice the heat is when you run things the S10+ would struggle to handle.
The problem I was actually pointing out was comparing the 865+ and 888. The 865+ clocks higher than the 888, but the 888 runs significantly hotter. Running the same games on an 865+ and an 888 will yield higher temperature from the 888, but only marginal performance improvement. The question is what happens when you try to run the 888 at its full capabilities?
Click to expand...
Click to collapse
Again its hot but NOT overheating. Anyways where I messed up in replying here is I thought I was in the Fold thread and NOT Flip. LOL. The Fold with its bigger chassis probably handles the heat a lot better. My daughter has the Flip and has not mentioned that it gets hot, but I now want to play with it.
And a side note I have been doing mobile device repair since the iphone was released and have worked on every snapdragon phone pretty much ever made.
JayRolla said:
Again its hot but NOT overheating. Anyways where I messed up in replying here is I thought I was in the Fold thread and NOT Flip. LOL. The Fold with its bigger chassis probably handles the heat a lot better. My daughter has the Flip and has not mentioned that it gets hot, but I now want to play with it.
And a side note I have been doing mobile device repair since the iphone was released and have worked on every snapdragon phone pretty much ever made.
Click to expand...
Click to collapse
I never said it was overheating, either. I said it has a much higher potential to overheat and environmental temperature will have a greater impact on it.
I've also been repairing phones since long before the iPhone, so I apologize that those credentials don't have more weight.
In all honesty, this 888 is much tamer than some. Take a look at the ROG Phone 5. You can melt an igloo running YouTube with that.
twistedumbrella said:
In all honesty, this 888 is much tamer than some. Take a look at the ROG Phone 5. You can melt an igloo running YouTube with that.
Click to expand...
Click to collapse
Anyone had the LG G Flex 2 with the 810? That thing got hot just being on the homescreen.
The Flip3 is nowhere near that and a non-issue heat wise for me. Especially now that I have the strap case where I don't notice it at all.
M4-NOOB said:
Anyone had the LG G Flex 2 with the 810? That thing got hot just being on the homescreen.
The Flip3 is nowhere near that and a non-issue heat wise for me. Especially now that I have the strap case where I don't notice it at all.
Click to expand...
Click to collapse
I honestly think Samsung did some behind the scenes limitations on it. The ROG Phone 5 and Z Flip 3 are both the 888, but there is a much larger performance gap than just a few minor software optimizations.
twistedumbrella said:
I honestly think Samsung did some behind the scenes limitations on it. The ROG Phone 5 and Z Flip 3 are both the 888, but there is a much larger performance gap than just a few minor software optimizations.
Click to expand...
Click to collapse
Yes samsung limit big core and prime core even when gaming pubg , wild rift , mobile legend. I jist finish an hour game pubg and see the result of big core and prime core even when running demanding game. Samsung does not fully utilize big and prime core. Most of its task use small core. Big and prime deep sleep very deep. Gpu will not run over 500mhz
I have the same issue when using Android Auto with it. Overheats to the point that music starts stuttering and then AA just shuts off completely. T-Mobile is sending me a replacement but the lady at the store said her friend had the exact same issue. I don't game or do anything that would require the SD888 to go nuts.
i got mine to overheat after a long video call with usb power on a somewhat hot day, had to close camera and give it a sec to cool down.
honestly its not terrible, but it does heat up more than regular phones. during regular use its fine imo
Hello everyone,
This thread will be used as a hub where I share some discoveries/observations which I stumbled upon mostly during working on my kernel projects.
I´ll clone the same thread over to the Pixel 7 forum as well. So without much further ado let´s just dive right into it.
A year ago everyone was excited for the Google SoC called Tensor 1 called GS101. One year later there is Tensor 2 called GS201.
I suggest to read about the differences, updated modem, ISP, TPU and GPU in various tech related articles.
Here´s a table so everyone gets up to speed on cores used, max freqs and other details:
View attachment 5744229
But how does that translate on the devices?
There were quite a few rumors before the actual release of the Tensor 2 SoC being manufactured on 4nm Samsung node instead of 5nm. However that was just wild speculation and unfortunately turned out to be not accurate. Tensor G2 is still manufactured in the 5nm process as confirmed by Google. This was quite a negative surprise to myself, as I don´t have good experiences from SD888 that´s also being manufactured in Samsung 5nm node and is quite a hot chip. While the switch to Samsung 4nm node, wasn´t all that great either (check sd8 gen1 on samsung 4nm vs sd8+ gen1 on tsmc 4nm) it would still have been an improvement.
While I was very excited for the Tensor G1 when the Pixel 6 devices launched, that excitement ebbed down the work I worked on the Pixel 6 series. The more I learned about the source, the more I stumbled upon Exynos driver over exynos driver, some are just left exactly like on Exynos device, some were "re-branded" by Google. Some Google did customize, but most of the drivers are just very much Exynos.
So all in all the following excerpt from Andrei Frumusanu´s article here sums it up pretty fitting:
While Google doesn’t appear to want to talk about it, the chip very clearly has provenance as a collaboration between Google and Samsung, and has a large amount of its roots in Samsung Exynos SoC architectures.
Click to expand...
Click to collapse
The same is true for the Tensor 2, despite minor upgrades there. As I learned over the time, Tensor shares a lot of Exynos characteristics, one of those is performance vs thermals as hinted by in the linked Anandtech article. So let´s just jump into that first topic I want to cover.
I will cover more topics, those will also be probably interconnected to each other, but we have to start somewhere.
Thermals, Thermal Ceiling, Exynos Roots and Maximal Performance
To start things of: Thermals is a term that actually sums up a few mechanisms. Lets split this into two main areas.
The thermal ceiling (let´s call it that) that´s being implemented in the kernel, as the maximal temperature the SoC is allowed to be operated at.
The thermal-hal uses combined sensors, also virtual sensor, and restricts different subsystems, based on the temperature of those sensor. Those can be called skin temps, shell-temps, battery temperature, modem temperature etc.
First let´s explore the thermal ceiling on the two SoCs:
GS101 on Pixel 6 devices is allowed to be operated at 90°C. GS 201 on Pixel 7 devices raised the thermal ceiling by 10°C to 100°C.
If changes to the internal design allowed them to raise this, without further increasing heatup of the device, or if they just applied changes to the thermal-hal to better keep this in check I don´t know at this moment.
Let´s get back to the Exynos characterstics. I talked to a few other developers I met along that way with Exynos "experience". Exynos SoCs reach the thermal ceiling extremely quickly, as I learned. This means, the SoC can´t keep its max CPU freqs for more than a few seconds without touching the thermal ceiling and getting restricted. This is in a way also the case for other SoCs, but Exynos is very extreme in this regard. But it´s just the characteristics of the SoC, like previously mentioned.
That means in turn: The thermal ceiling is setting the maximal performance allowed, to a great extent. If the thermal ceiling is raised, the maximum performance can be held longer.
Here´s a demonstration of this:
Pixel 6 Pro in its default configuration running at 90°C temp ceiling:
New video by freak 07
photos.app.goo.gl
Pixel 6 Pro with temperature ceiling raised to 100°C, instead of 90°C running at Pixel 7 Pro clocks
New video by freak 07
photos.app.goo.gl
That´s the Pixel 7 Pro, with the default configuration of 100°C
New video by freak 07
photos.app.goo.gl
Now what´s interesting, the big cores get the hottest at the quickest rate. Once the ceiling is reached, the max performance drops, as maximal performance will be restricted by restrict max cpu freqs.
That´s the case after a few seconds on both SoCs, in typcial Exynos fashion.
Let´s make the next connection:
Although I´m not necessarily a friend of benchmark apps, how does that change the results of a CPU oriented benchmark like Geekbench you might ask yourself. There are other benchmarks, but I want to keep this simple for now.
The answer is: The Pixel 6 Pro with GS101 gets pretty close to the results of the Pixel 7 Pro with GS201.
So for comparisons sake:
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
On the left Pixel 6 Pro in its default configuration running at 90°C temp ceiling.
On the right Pixel 6 Pro with temperature ceiling raised to 100°C, instead of 90°C running at Pixel 7 Pro clocks.
The kernel used was the same, no changes to anything that could impact performance.
The left screenshot above was taken from the Pixel 7 Pro review from XDA, while the right one was taken on my Pixel 7 Pro running my kernel.
Please don´t start benchmark contests now, It´s just for comparisons sake.
It makes sense for single-core to be less impacted, as single core benchmarks don´t put as much thermal pressure on the SoC -> not touching the thermal ceiling as much and therefore no cutback are applied.
Geekbench applies a series of short benchmarks to the device. Usually not longer than 3-8 seconds, which is ideal for a SoC like the Tensor. Short bursts with max performance, so it can run "nearly" without touching the thermal ceiling.
If a benchmark is structured differently, like the CPU stress test you will see QCOM SoCs holding their max-freqs for minutes, instead of seconds.
Well that´s the first part. More to come. I hope everyone enjoyed this little writeup so far.
I wish everyone a nice evening.
Thermal Ceiling/Maximal Performance - A comparison between QCOM Snapdragon and Tensor
Now you might ask yourself, how does QCOM´s Snapdragon behave in the little test we conducted above.
You can find the answer below.
For this a Zenfone 9 with the Snapdragon 8 + Gen 1 is used.
Pixel 6 Pro in its default configuration running at 90°C temp ceiling:
New video by freak 07
photos.app.goo.gl
Pixel 7 Pro, with the default configuration of 100°C:
New video by freak 07
photos.app.goo.gl
Asus Zenfone 9, with the default configuration of 110°C temp ceiling:
New video by freak 07
photos.app.goo.gl
As you can see the Zenfone 9 with SD 8+ Gen 1 can hold the max-freqs for minutes. It doesn´t touch the thermal ceiling when running under max load for a minute, while Tensor immediately scratches the ceiling.
I´m not a SoC expert and I think only engineers with insider knowledge know the exact reason why Exynos based SoCs behave that way. They just seem to work totally different in that regard.
Another point is, since the SoCs are different we can´t compare the temperatures one to one. There´s no way for me to know the exact placement of the temperature sensors, all I can say for sure is the SD 8+ Gen 1, does not touch the thermal ceiling in this test and there seems to be a lot of headway after one minute of maxed out CPU.
In the end the result will be the same. After a while the device will heat up and the thermal-hal will throttle the ZF9 back as well, as with only passive cooling that´s inevitable.
this one as well
this here too
and this one here as well
last one
Glad to have same phone as you good work
Freak07 said:
I hope everyone enjoyed this little writeup so far.
Click to expand...
Click to collapse
I am! This is very cool; thanks for doing this! Most of us (myself included) don't have the knowledge to do (or interest in doing) this ourselves, but enjoy learning more. Thank you for all that you do, both this writeup and your kernel (which I've used since the Pixel 3 days).
@Freak07 Interested in your thoughts on this CPU 'Power Efficiency' comparison. Seems a very well thought out review, although only compares the G1 Tensor.
updated the second post, with a comparison between QCOM´s Snapdragon and Google´s Tensor
Google never intended Tensor to be a flagship SoC, it was just meant to have bursts of flagship-level performance. I haven't had any complaints about the day to day or demand performance honestly, it's been WAY smoother than my OnePlus 9 with the SD888 that almost sent me back to Apple, but I will confirm that I ran the full 15 minute test when I first got the phone and it throttled to 78% near the end.
P7 also appears to stop charging, or charge at 500ma once the battery reaches 40c, which is a concern since high drain situations are usually the time where you want to have the device plugged in.
There's also not much headroom since normal temp alone when charging with no load is around 37c.
One thing I haven't noticed is throttling, this phone gets hot and is inefficient, but I haven't experienced any throttling on any day to day use.
My S22U is ice cold when playing media and quickly becomes a dumpster fire once you load up a game. The p7p runs hot when playing media and barely loses performance during gaming. It's weird.
Agenesis said:
The p7p runs hot when playing media and barely loses performance during gaming. It's weird.
Click to expand...
Click to collapse
What media were you running out of curiosity?
EtherealRemnant said:
Google never intended Tensor to be a flagship SoC, it was just meant to have bursts of flagship-level performance. I haven't had any complaints about the day to day or demand performance honestly, it's been WAY smoother than my OnePlus 9 with the SD888 that almost sent me back to Apple, but I will confirm that I ran the full 15 minute test when I first got the phone and it throttled to 78% near the end.
Click to expand...
Click to collapse
What really defines a flagship SoC?
For me tensor pretty much is a flagship SoC. Not necessarily ẃhen chasing the highest benchmark scores.
I´ll link my post from the p6 pro thread here.
Agenesis said:
P7 also appears to stop charging, or charge at 500ma once the battery reaches 40c, which is a concern since high drain situations are usually the time where you want to have the device plugged in.
There's also not much headroom since normal temp alone when charging with no load is around 37c.
One thing I haven't noticed is throttling, this phone gets hot and is inefficient, but I haven't experienced any throttling on any day to day use.
My S22U is ice cold when playing media and quickly becomes a dumpster fire once you load up a game. The p7p runs hot when playing media and barely loses performance during gaming. It's weird.
Click to expand...
Click to collapse
Charging while the device is hot is not good for battery health. Don´t charge while your device is under load as well. Both are detrimental to battery health.
Charging is done, so it usally reaches close to, but doesn´t cross 40°C if you don´t touch the phone.
I don´t have that at all. Watching streams, youtube videos the device doesn´t get hot. Maybe a bit warm, especially if I load a few apps in between, but not hot.
I don´t game much though, so I can´t comment much on this matter.
I played a bit of this game and the device didn´t get hot as well. Though I guess you can´t really call that a demanding game. Maybe it´s also optimized quite well for android.
i dont understand obsession on benchmarks for phones.
these are not PCs lol
as long as the phone does not lag or slow down when doing everything you need then its fine
Freak07 said:
What really defines a flagship SoC?
For me tensor pretty much is a flagship SoC. Not necessarily ẃhen chasing the highest benchmark scores.
I´ll link my post from the p6 pro thread here.
Charging while the device is hot is not good for battery health. Don´t charge while your device is under load as well. Both are detrimental to battery health.
Charging is done, so it usally reaches close to, but doesn´t cross 40°C if you don´t touch the phone.
I don´t have that at all. Watching streams, youtube videos the device doesn´t get hot. Maybe a bit warm, especially if I load a few apps in between, but not hot.
I don´t game much though, so I can´t comment much on this matter.
I played a bit of this game and the device didn´t get hot as well. Though I guess you can´t really call that a demanding game. Maybe it´s also optimized quite well for android.
Click to expand...
Click to collapse
The Tensor is definitely meant to be upper midrange with bursty performance rather than a flagship with steady high performance. I mean Google could have gone to Samsung 4nm and chose not to while being totally aware of the peak performance issues with the first gen Tensor and even bumped the clocks so it seems to me they aren't really unhappy the position of the Tensor chip relative to everything else. It's going to be a distinctly midrange chip compared to the 2023 flagships if the rumors about the gains Qualcomm got from switching to TSMC end up panning out.
EtherealRemnant said:
The Tensor is definitely meant to be upper midrange with bursty performance rather than a flagship with steady high performance. I mean Google could have gone to Samsung 4nm and chose not to while being totally aware of the peak performance issues with the first gen Tensor and even bumped the clocks so it seems to me they aren't really unhappy the position of the Tensor chip relative to everything else. It's going to be a distinctly midrange chip compared to the 2023 flagships if the rumors about the gains Qualcomm got from switching to TSMC end up panning out.
Click to expand...
Click to collapse
I have a theory I've mentioned before that since Google depends on Samsung for so much, you have to consider that Samsung doesn't necessarily make the best of the best available for Google to purchase at a reasonable price since Google's products effectively compete with Samsung's own, and Samsung has Google more "over the barrel" than Google does Samsung.
roirraW edor ehT said:
I have a theory I've mentioned before that since Google depends on Samsung for so much, you have to consider that Samsung doesn't necessarily make the best of the best available for Google to purchase at a reasonable price since Google's products effectively compete with Samsung's own, and Samsung has Google more "over the barrel" than Google does Samsung.
Click to expand...
Click to collapse
This is certainly possible. I also think that Google is pretty happy with Tensor overall. They have never marketed this phone as the top performing device but the performance it does have is pretty solid. Until I installed the beta, I hadn't even seen any random lag, which is amazing to me coming from the mess that was the OnePlus 9.
Thought this might be a good place to share this...
Chip chat! And not the cool ranch kind. In the fourth episode of the #MadeByGoogle podcast, our host Rachid Finge speaks with Monika Gupta, Senior Director of Product Management for Google Silicon Teams, about the Tensor G2 chip in Pixel 7 and Pixel 7 Pro. Monika and Rachid discuss Google’s AI first approach, how Tensor got its name, its role in our favourite speech and camera features, and why classical benchmarks for chips don’t always tell the whole story.
Episode 04 – Chip Chat
Chip chat! And not the cool ranch kind. In the fourth episode of the #MadeByGoogle podcast, our host Rachid Finge speaks with Monika Gupta, Senior Director of Product Management for Google Silicon Teams, about the Tensor G2 chip in Pixel 7 and Pixel 7 Pro. Monika and Rachid discuss Google’s AI...
made-by-google-podcast.podigee.io
This thread still alive/relevant? 'Cause I've been looking for information on how to access the TPU on the Tensor chip and didn't find anything
My dream is to be able some day to run LLaMA on Tensor's TPU via ggml/llama.cpp - but for that you'd need to know how to access the TPU directly and that doesn't seem to be possible in any way... or may be I missed something?