Manual Change log URL : 
\
https://chromium.googlesource.com/chromium/src/+log/56.0.2915.0..56.0.2916.0?pretty=fuller&n=10000

Attaching the screenshot of Google chrome and canary CPU Usage , whenever the issue is reproduced.

Thanks !

Deleted: CPU Usage.png 226 KB

	CPU Usage.png 226 KB View Download

Deleted: CPU Usage_1.png 364 KB

	CPU Usage_1.png 364 KB View Download

Had an offline chat with guidou@, seems like the above change is not related. I am trying to run per-revision bisect to see if that helps.

+erikchen for mac CPU issues. 

Seems like i am still missing something here. *CPU never turns off* means even after stop playing the videos in different tabs? or existing the chrome browser/processes? I have tested this behavior for Mac OS X 10.11.6 (Mac Book Air, 13-inch, Mid 2012)on Latest Canary#56.0.2919.0, Dev#56.0.2914.3, Stable#54.0.2840.98 and the above bad build#56.0.2915.0 - CPU fans are getting turned off only if i stop playing videos or exiting the browser. Otherwise fans are still running (and making considerable noise) across all the above chrome versions.

kerz@, Just wondering are you still seeing this issue on Latest Canary#56.0.2919.0 consistently?

erikchen@, I am not able to find the possible suspect from the change log (https://chromium.googlesource.com/chromium/src/+log/56.0.2915.0..56.0.2916.0?pretty=fuller&n=10000). Would you be able to find something related?

At the moment removing RBD label and please feel free to add it back in case there is a clear repro.

Thank you!

*CPU fan never turns off*

I'm suffering this too. I'm on 56.0.2919.0.

It started today. A simple Youtube video consumes most of my MBP CPU and the fans start spinning like crazy. Even simple websites consume as much as 50% of my CPU and keep the temps around 70ºC which is crazy.

If you have this problem, please record a trace and upload it here:
https://www.chromium.org/developers/how-tos/submitting-a-performance-bug

I've tried to record a trace a few times but it crashes Chrome every time.

Is there anything else I can do to help you guys?

Can you go to chrome://crashes and let us know the crash ids?

The crashes from today don't appear there...

I've tried recording a new trace and after it reaches 100% it says something like "2 out of 2" for a minute or so until the tab dies and Chrome asks me to kill it or wait.

Here you can see CPU usage only with Gmail and Youtube playing a video.

http://i.imgur.com/4IkrnXD.jpg

1 is when I'm playing a video with gmail
2 is when recording the trace
3 is after recording the trace

BTW I'm on macOS 10.11.6

Oh - can you stop the trace really early? It doesn't need to hit 100%. Just start the crash, switch to the gmail/youtube tab, switch back, and then stop the trace.

I was able to get the trace. Here it is.

Deleted: trace_chrome_tracing.json.gz 4.2 MB

trace_chrome_tracing.json.gz 4.2 MB Download

Thanks!

In this trace, the crbug.com renderer is going crazy. Every 0.1ms, there's a callback from 
"""
src_file	
"../../third_party/WebKit/Source/platform/mac/ScrollAnimatorMac.mm"
src_func	
"startScrollbarPaintTimer"
"""

There may be other issues, but this by far jumps out the most. +bokan, who based on the history of this file, appears to be an expert on scrolling/painting

I am also seeing this issue (CPU usage is running crazy!) on my home n/w with Mac OS X 11.6 having very few tabs as Gmail, crbug and few other internal sites. Not playing any video as such. Here is the chrome://tracing for reference: https://drive.google.com/a/google.com/file/d/0B-vMgxWkGDE2eXBNVTZOWDhQbGc/view?usp=sharing

PS: I used this machine whole day at work and didn't notice this behavior.

Thank you!

Tried again the hasbisect-per-revision but Issue is very inconsistent to provide the reliable tool bisect result. Tested this on MacBook Air OS 10.11.6.

as per C#15, based on the code search on 'ScrollAnimatorMac.mm' seeing most recent change by bokan@ https://codereview.chromium.org/2467693002.

bokan@: Could you please take a look at this and help in further investigation if this could be related to changes from Issue 662402(Mac specific).

Appreciate your help!

This was likely caused by https://codereview.chromium.org/2454913003/ and was reverted recently (though it would still be in 56.0.2919.0) - this is probably a dup of 664809. Could anyone seeing this try 56.0.2920.0 today and report if it's any better?

I opened Chrome Canary with 56.0.2920.0 (it didn't update since it's the same version?) and CPU usage was lower than yesterday although much higher than stable channel.

Here you can clearly see when I opened Canary.

Deleted: Screen Shot 2016-11-15 at 9.38.34 AM.png 85.5 KB

	Screen Shot 2016-11-15 at 9.38.34 AM.png 85.5 KB View Download

Is that when playing video? Or simply scrolling pages? Could you please attach a new trace for comparison?

4 tabs opened:

- tracing
- youtube playing a video
- crbugs.com
- gmail

Deleted: Screen Shot 2016-11-15 at 10.26.01 AM.png 62.5 KB

	Screen Shot 2016-11-15 at 10.26.01 AM.png 62.5 KB View Download

Deleted: trace_Tue_Nov_15_2016_10.26.13_AM.json.gz 10.0 MB

trace_Tue_Nov_15_2016_10.26.13_AM.json.gz 10.0 MB Download

Comparing the new trace and old, the original problem is gone - the scrollbar paint timer in an infinite firing loop. The CPU usage in the latest trace appears to mainly come from the gmail tab which is running a large amount of JS. This can be just back luck as it decides to do GC or other background tasks, but CPU usage went down towards the end. I'd also expect some CPU usage from a playing video. The original report had to do with CPU saturation while the browser was idle and that's fixed so I'm duping into that bug. Feel free to open a new bug if you're still seeing noticeable differences between stable and canary.

There is still a pretty big difference between stable and canary which wasn't there a day or two ago. If you look at my last screenshot the difference is abismal, and I had many more open tabs in stable than canary.

Ok, sorry to keep asking you for info but I need an equivalent comparison to tell the difference: could you file a new bug with a trace from both stable and canary on the same page, doing roughly the same thing. As it is, the trace above doesn't show anything obviously wrong but there's a lot going on. If there really is a big difference on Canary it should show up in a comparison. Thanks!

No problem!

I'll file a new bug later today.

Issue 697042 has been merged into this issue.