Root SetCPU Widget, Launcher testing, Heat, CPU

[jtbnet]

Maybe this is Not New news to folks but I had not seen it mentioned here or in other forums... The recent SetCPU widget on a home screen can seem to get into some kind of race condition where procs, launcher(x), Android System, and SetCPU seem to start eating up CPU cycles and my CPU Temp will rise over time at least as much as 60 deg F. which is pretty dramatic in heat and battery life...

I have been comparing features and performance from my 3 favorite Launchers for the last couple of weeks... usually 2-3 days solid on each launcher running as close to identical layouts and widgets as possible... I have had SetCPU set to 400/1100 ondemand mostly and when I start either LauncherPro, or ADW and let it get to setup and idle my CPU temp will be as low as 60F... over time (hours, over nite... etc...) I notice it creeping up till it's sitting at 130+ and when I started watching snapshots of procs with OSmonitor I see the above 3 procs all seeming to incrementally increase in percentage of CPU use with the temp rise....

I had the SetCPU widget set to active refresh every 5 sec. and for the first couple hours you wouldn't really notice much impact from it... but over time it definitely had increasing (like racing procs..) impact .... yesterday I finally decided to remove the SetCPU widget from my home screen and today my temp has been basically rock solid at bottom end of what I've been seeing... (60-80F)...

Moral of story... be careful running SetCPU Widget... at least using active updates of current CPU speed... (didn't seem make much sense to me to use without active updates as just tells me what my default profile settings are...)

 

[inssane]

Most people don't do such an aggressive refresh because it does create more heat and sucks battery. I have temp monitor on my top bar refresh every 60 seconds and I use SetCPU for the profiles mainly.
Most importantly, the kernel makes a difference as well...

 

[jtbnet]

I'm probably not making my notice/issue very clear here... I'm not trying to explain an issue/bug that may be able to be dealt with here... but an interesting 'occurrance' of a VERY simple process's interaction/Race with just a couple other processes which can cause a runaway race condition... which can even go as far as to trigger the Fail Safe profile on SetCPU, no less dramatically impact Temp and Battery life.

I agree 5 Second refresh on most display widgets is agressive... but this is NOT the root of the issue as I run Agressive refreshes on multiple processes, especially when I am trying to profile system behavior. In this case I have TempMon, SetCPU Widget, and OSMonitor all refreshing at their most agressive 5 second interval and the typical CPU overhead at 400/1100 is maintaining 3% @ 400 total CPU usage which runs Extremely cool (using Chevy's latest 1.2 LV kernel)...

The Race/runaway does not happen immediately, and may even not start to occur for 3 or more hours, but once it starts, I've seen the CPU usage of Just the 3 processes creep up to 86% @ 1100... this occurs with All 3 Launchers I have been testing, thus the 3 process interaction is between;

Android system
Launcher(X)
SetCPU

Another point here is that the 5 second refresh does not actually trigger the issue by itself... a less frequent refresh may make this occur less frequently, but i have seen it occur with a 30 sec. refresh on SetCPU... thus the fact it sometimes takes a long time to start occurring...

If you weren't watching for this particular set of events the phone might just all of a sudden seem to lag dramatically, or battery will noticibly Plummet, and a simple reboot makes it go away for a while... Also occasionally the situation may just clear itself after some time or additional event triggers that might be unlinking events of the race thus breaking the seeming forever loop running... you just come back and you battery has drained considerably with nothing unusual now seeming to be going on...

This may not seem very exciting overall to many folks but having developed and supported various Unix variants over the last 30 years interesting race conditions like this occur every once in a while and can just be a challenge at times to track down, so I find them interesting... and I figured after seeing so many posts about strange lagging, and unusual battery drain that it was worth passing along at least as much as I've come to understand of this particular set of events...

The short of it all is it seems to me to be safest to Not use the active refresh of the SetCPU widget... I have had the widget back on my screen with the Passive option checked for close to 24 hrs now and my phone has used little battery and is running as cool as a cucumber....

 

[Fabolous]

You should let the developer of SetCPU know about this.

I've been using the passive refresh, so I haven't experienced this issue, but it's good to know about. Thanks!

 

[jtbnet]

You should let the developer of SetCPU know about this...

I couldn't find a direct email address to Mike so I did PM him over at xda-developers with a link to this thread so hopefully he'll see the PM...