Help 100% CPU usage, but it doesn't add up

[Palmetto Fellow]

Last night, my CPU usage jumped to 100% and stayed there for several minutes

 

[Mr. Ed]

I really wish I had an answer...there have been a few similar posts in the past few weeks and no one has figured it out that I am aware of

 

[Palmetto Fellow]

it happened again to me in the parking lot of a walmart yesterday except this time I saw the cpu usage . I was trying to pull a website from newegg when it just froze up and cpu usage went to 100 percent. the internet task was at about 85 percent and I was able to force stop it.

the 3 g coverage was pretty weak and I think that had something to do with it. I will do more experiments near that walmart to see if I can replicate this.

EDIT:

This is not a related issue. I was able to account for the 100% usage with Android System Monitor, and after closing the Internet task, the CPU use returned to normal. After learning more about my issue I have come to disqualify this occurance as relevent.

 

[Palmetto Fellow]

It happened again.

Today, I was watching a video using the Break app, and I decided to fast forward. It slowed down a bit when I did this, and within a minute of trying to move the timeline slider back and forth, I saw my CPU jump to 100%.

Using OS Monitor, I only showed 34% usage, but this time, Android System Monitor acknowledged 100% usage, but was unable to account for where it was being used.

I killed the Break app, but that did nothing to help. I ultimately reset the phone using both volume buttons and pwr.

After the reboot, i went right back to the same video and moved the timeline slider all over the place. I was unable to recreate this issue.

 

[Palmetto Fellow]

OK, so this happened yet again. I've been using the Break app more and more trying to recreate a problem.

And it happened while using Break. I was in 4G mode watching a video when I hit the back button. It delayed for a bit, but eventually allowed me to go back. When I got to the home screen, I saw 100% usage.

I opened Android System Info and again, the usage was not accounted for.

I closed the Break app, and that still didn't help. This time however, turning off the 4G did fix the problem. Yesterday when I did this turning the 3G and 4G radios off did not help at all.

Is there something that may provide a valuable log of what's going on?

EDIT:

My current theory is that the data signal goes in and out, and if that causes the 4G to switch to 3G, then my IP address will change, which may throw off the Break app. However, when I got back to the homescreen, the 4G logo was still there, and hadn't changed to 3G. I guess I could have missed the switch, and by the time I got to the homescreen, it went back to 4G...

Is this feasible? Would it affect any streaming source if I switch back and forth between 3G and 4G?

EDIT AGAIN:

Happened again...this time, I was trying to make it happen using Break. I stopped that, and someone asked me something about the weather...so I opened the Weather Channel app. While I was using the Weather Channel app, my CPU went to 100% and is still there. The Break app was not closed, just running in the background. I closed both it and the Weather Channel app. and Break app. I toggled the 3G and 4G off and on...nothing has fixed it yet.

Another interesting fact...OSMonitor does not pickup on the 100% CPU usage like Android System info and SeePU does. Does this provide any insight on what's going on? In other words, where do these three apps get their CPU usage statistics from, and why is one of them giving incorrect usage?

 

[Palmetto Fellow]

OK, I am almost certain that 4G is causing my 100% CPU issue.

Today, I was using the Break app with 3G but that was too slow for me, so I went and turned on 4G.

In the 30 seconds it takes for 4G to turn on completely, I received a call and began talking. During the call, I noticed my CPU usage at 100% again. After the call ended, it didn't drop at all. I had to reboot.

I found another thread where someone is having lag issues with 4G.

http://androidforums.com/evo-4g-sup...s-evo-very-laggy-please-help.html#post1812854

Can ANYONE in this forum offer advice other than to contact Sprint? I am thinking about a factory reset this weekend...


EDIT:

I went ahead and uninstalled the Break app as well. The two recurring themes seem to be 4G and Break. Break isn't always in the foreground, but it's running to some degree each time (I think) so I got rid of it for now.

Next week when I am back in Charlotte, I'll keep it in 4G as much as possible.

 

[Mr. Ed]

have you tried a 4g toggle from the market vs the htc toggle?

 

[Palmetto Fellow]

have you tried a 4g toggle from the market vs the htc toggle?

For now, I am going to do everything I can in 4G when possible...except run the Break app. If it lasts a whole week, I'll install Beak again, and use it. If the CPU gets screwy shortly thereafter, I'll know to uninstall Beak, or run it only in 3G.

I'll check into the toggle if this testing gets me nowhere. What if I switch 4G on directly from the Wireless settings? Would that be an effective way of ruling out my toggle?

 

[Vincent Law]

It's the 4G connection itself, not the App. This has been widely reported, it starts 100% usage once enabled, or when reconnecting to it. Rebooting is often the only solution. It's not really CPU usage, it's I/O wait.

 

[Palmetto Fellow]

It's the 4G connection itself, not the App. This has been widely reported, it starts 100% usage once enabled, or when reconnecting to it. Rebooting is often the only solution. It's not really CPU usage, it's I/O wait.

If that's the case, someone needs to report it more often to HTC. I called them about this issue last week, and the rep looked into it a bit and told me that there isn't a lot of reporting about 4G causing these issues.

Either they are in denial, or people aren't reporting it to HTC, only on forums or whatnot...

 

[Vincent Law]

More like many people don't realize the cause of it, because it's not exactly like the phone screams "HEY YOUR PHONE IS BEING SLOW BECAUSE OF A 4G BUG".

Also, I am running the latest leaked ROM and can no longer produce the issue, so perhaps they're already fixing it.

 

[Mr. Ed]

In my case connecting via settings or sense toggles resulted in lag. I connected several times with jasons toggle and had no lag

It works for me...may not for others. There is also a link over at xda about changing some 4g connection settings....I will try and find it when I get home.



Vincent....as I said in the other thread.....if you are going to push your theory at least back it up. You made it quite clear you have issues with jason and his business practices concerning the toggle give away. That is no excuse for discouraging others from what could be a solution to their problem.

If that's the case, someone needs to report it more often to HTC. I called them about this issue last week, and the rep looked into it a bit and told me that there isn't a lot of reporting about 4G causing these issues.

Either they are in denial, or people aren't reporting it to HTC, only on forums or whatnot...

 

[Vincent Law]

Vincent....as I said in the other thread.....if you are going to push your theory at least back it up. You made it quite clear you have issues with jason and his business practices concerning the toggle give away. That is no excuse for discouraging others from what could be a solution to their problem.

Because you're telling people it's a solution to a problem when it 100% isn't. Even the developer of the 4G widget you're talking about never said anything about it fixing the issue. I have no problems with the widget. The fact that you haven't stumbled upon it yet either means:

A. You're running the latest leaked ROM which may have fixed it.
B. Luck.
C. Your phone NEVER exhibited the issue, but I doubt that, as I've yet to hear of anyone's phone not have the issue.

There's nothing in that widget that would be capable of NOT causing the issue. Seriously, it would be MAGIC if that happened, because the widget doesn't even have the access to the files necessary to fix the issue if it wanted to.

Remember, this issue is NOT something you can repeat on demand. That is why it's difficult to simply state "Do X, Y, Z". The issue happens when turning on, or if it's already on, when connecting to 4G. It does not happen all the time, and especially after a reboot.

There's been some debate on XDA about it being related to a bad tower state (since it happens more often in areas with 4G being added). So it may be that the tower you're using is the cause and solution as well.

Final Edit: The burden of proof here is on you. You say it's a fix to the problem, explain how it could have fixed the issue. I have common sense, and technical reasoning for why it can't. You need to explain why it could.

 

[OfTheDamned]

Please stay on the current topic of this thread. If you have anything to debate that is off topic, I suggest you take it to PM instead of hijacking other's threads.

 

[Vincent Law]

Thanks mod, I've told him to PM me if he has an issue with me a while ago, he seems to refuse.

In any case, I think this thread's been wrapped up anyway.

 

[Palmetto Fellow]

So far today, I've been using 4G for a few hours.

My CPU usage jumps to 100% to load the pages, but then goes back to normal as it should once the page is loaded. Also, when checking CPU load, the itemized list shows 80% or more to the Internet.

This is not what I get when I have the crash issue described in the OP. When the CPU goes haywire (I/O wait as described by Vincent), I don't get any itemized tasks which account for the usage.

As mentioned, I am going to refrain from using the Break app. So far, I've only been using the stock internet browser to pull lots of graphically intense pages with lots of hot linked images and animated GIFs. So far, the CPU jumps to 100% and stays there until I close the Internet app...then everything's back to normal.

I have yet to recreate my isse, but will continue to try throughout the day. I also wonder...could it be that when an app uses 4G there are issues? Since I've been using the stock Internet browser so far...maybe that's why it's working. Is there some sort of bridge that an app needs to utilize in order to grab the internet connection?

Maybe there is something wrong with 3rd party apps using 4G?

Can you confirm that this happens with 4G on and using the stock browser? Has this problem been reported in this manner?

 

[Vincent Law]

The 4G issue will actually happen even if there isn't 4G signal, as long as it's turned on, and with no apps running.

Also, you may try the System Panel app, and turn on "System Processes" in the settings. When the 4G bug is present, the "System" process jumps up to ~30-40%, and doesn't go down. The whole CPU will be pegged at 100%. It's a way of determining that we're on the same thing.

 

[Palmetto Fellow]

The 4G issue will actually happen even if there isn't 4G signal, as long as it's turned on, and with no apps running.

Also, you may try the System Panel app, and turn on "System Processes" in the settings. When the 4G bug is present, the "System" process jumps up to ~30-40%, and doesn't go down. The whole CPU will be pegged at 100%. It's a way of determining that we're on the same thing.

OK, I've installed System Panel Lite. That's a pretty sweet looking app, thanks for sharing.

Also, how do I turn on 4G without signal? Do you mean just having it on while I am in 3G mode (but out of 4G range).

Or should I turn on 4G and turn 3G off? This will force the phone to use 4G or nothing.

BTW, if the signal is not needed, then it's probably not a tower issue right?

 

[Vincent Law]

I meant that it doesn't have 4G signal.

It's not set in stone the cause, but for example, I am within range of a 4G tower, but that tower doesn't always have working 4G (as they're still building it up in my area). Thus, sometimes it appears to have no signal, other times it does. It's all theory though as to the cause, other than it's 4G related. But I don't even have (nor ever did) the app you describe, so I doubt it's directly related to it. The other common theory was widgets as mentioned, but that was disproven since it can happen even without any widgets.

Next time the CPU issue pops up, check that system process and tell me what you find, alright?

 

[Palmetto Fellow]

Next time the CPU issue pops up, check that system process and tell me what you find, alright?

It happened during lunch, before I was able to get my stuff together with the new app.

I was streaming Pandora this time (another non-native app) and it went to 100%. Oddly enough, going into airplane mode fixed it. This has never worked before. I then went back to 4G and started Pandora back. Within a few seconds it went to 100% and airplane mode didn't fix it this time.

System Panel didn't show 30% for system (wasn't even close) but as mentioned, I haven't really used it at all and wasn't sure exactly what to look for. I did notice some weird things with System Panel though. On the main page, it showed 100% CPU usage. But when you click on the 3 pie charts at the top and it goes to a zoomed view of those charts, the CPU usage was never over 50%. The itemization for the CPU pie chart showed nearly 78% Idle most of the time. This behavior mimicks that of Android System Monitor where it shows 100% CPU usage as a whole, but when you ask it to show you how the CPU is being used, it can't. Is this phenomenon described as the I/O Wait? Is that like a phantom CPU thief?

Back to your point, I didn't see System usage similar to what you describe, but it was my first time with that app. Next time, I'll be sure to check it further.

I use 4G in Charlotte. This was one of Clear's earlier markets...right after Atlanta, and I don't think my location in Charlotte is a place for a new tower. It's a pretty busy section of town, and would have been a part of the original rollout most likely.

Can you direct me to some other discussions of this problem? Is there a name for this phenomenon I can google?

 

[Vincent Law]

Interesting, as system process always spikes when it does it for me.

And yes, that phenomenon is the I/O wait issue. Basically, the CPU isn't really /busy/, it's stuck waiting on RAM/disk response. Since this is an in-order processor (all smartphone CPUs are), they have a hard time working around lots of I/O wait, and it basically brings the phone to its knees. It'll say the CPU usage is high, because it can't be used for much else. Note that this isn't a problem with the RAM/disk, it's the fact that some code is requesting some data from RAM/disk repeatedly, VERY quickly, over and over, which overloads it. I read once that someone said it had to do with the 4G device in Android being constantly queried (Linux makes devices accessible as files in the file system), so if that is true, that would be I/O wait. I haven't seen anyone do anything indepth on this though, because it's hard to consistently reproduce, and when it does happen, getting anything done to check it is sloooow.

Animations and such will sometimes appear smooth thanks to the task scheduler, but anything requiring I/O (like loading up another app, etc) will take ages.

You'll probably find a lot of discussions on this by simply searching for 4G bug.

 

[Palmetto Fellow]

Searching for "4G bug" didn't work well for me. The fact that the EVO's name has 4G in it leads to lots of bad hits. I can spend more time looking itno this later.

BTW, if you search for

"EVO 4G Lag I/O wait" a link back to this thread is the first hit. This is a bit worrisome, as I had hoped more people were suffering from this.

Is this supposed to affect all EVOs, or some EVOs? If all, I would surely think that there would be more noise about this.

Also, what have you found with regards to logging services? Is there anything that can actually document what is going on with those I/O wait requests? You mentioned that part of the problem is the randomness of this issue.

My experience has been that I can recreate this problem every other day if I try hard enough. I'd like to run something that may capture what's going on.

EDIT:

This reminds me, I have a message log, but I am not 100% I made it during this problem (though I think I did). I think OS Monitor allows you to export a message log...and this is the output from that log:

10/22/2010 14:19:26 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9215 objects / 524080 bytes in 40ms
10/22/2010 14:19:27 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9218 objects / 525048 bytes in 39ms
10/22/2010 14:19:27 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9212 objects / 523672 bytes in 40ms
10/22/2010 14:19:28 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9247 objects / 525912 bytes in 41ms
10/22/2010 14:19:29 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9209 objects / 522624 bytes in 62ms
10/22/2010 14:19:30 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9210 objects / 524128 bytes in 39ms
10/22/2010 14:19:30 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9209 objects / 523872 bytes in 39ms
10/22/2010 14:19:31 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9218 objects / 524768 bytes in 40ms


--SNIP--

10/22/2010 14:19:39 [WARNING] IInputConnectionWrapper(959) showStatusIcon on inactive InputConnection
10/22/2010 14:19:40 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9206 objects / 524800 bytes in 50ms


--SNIP--

10/22/2010 14:19:47 [DEBUG] RequestQueue(959) setProxyConfig mProxyHost=null
10/22/2010 14:19:48 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9188 objects / 526296 bytes in 40ms

--SNIP--

10/22/2010 14:21:42 [DEBUG] RequestQueue(959) setProxyConfig mProxyHost=null
10/22/2010 14:21:43 [DEBUG] dalvikvm(959) GC_FOR_MALLOC freed 9233 objects / 526720 bytes in 38ms







It was a 3 minute log I think, with hundreds of entries…I just copied the first several to give you an idea of what it’s saying. Does this mean anything to you? Also, I went back and noticed a couple of lines that were different from 99% of the other messages, and made them bold.

 

[Vincent Law]

The MALLOC messages are ignorable.

I haven't done any looking at the logs personally, and as I'm still unable to reproduce the issue on the new leaked ROM I'm running, I can't look myself.

To have something that would peg the CPU in 100% I/O wait would require log messages on the order of 100's per second.

 

[Mr. Ed]

Palmetto are you rooted? Vincent mentioned that he has updated to a newer rom with the leaked htc update and isn't seeing the issue .....if you are rooted it is definitely something to look into, though as he also mentioned the lag can be present hit or miss

 

[Vincent Law]

I am also rooted, rooting would have no influence on the 4G service or device. People have reported the 4G lag on both rooted and unrooted ROMs.