Closed Bug 1438510 Opened 7 years ago Closed 7 years ago

Very high CPU usage on Windows 8.1 Pro (profile + video included)

Categories

(Firefox :: Untriaged, defect)

60 Branch
defect
Not set
normal

Tracking

()

RESOLVED INCOMPLETE

People

(Reporter: david.hellfeuer, Unassigned)

Details

User Agent: Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:60.0) Gecko/20100101 Firefox/60.0 Build ID: 20180214224814 Steps to reproduce: I'm filing this following up on the call for action on Reddit: https://www.reddit.com/r/firefox/comments/7knnn4/firefox_quantum_is_eating_your_cpu_help_us_debug/dt86jbf/ Firefox: Today's Nightly 60.0a1 (2018-02-14) (64-bit) System: Windows 8.1 Pro (64-bit) CPU: Intel Core i7-4600U @ 2.10GHz (4 CPUs) GPU: Intel HD Graphics Family, Approx. Total Memory 1792 MB RAM: 8 GB General: DELL laptop, System Model: Latitude E7240, BIOS: A08, DirectX version 11 Steps to reproduce: 1) Start Firefox Nightly 2) (Optional) Go to news site http://www.zeit.de 3) (Optional) Go to sports site http://www.kicker.de 4) (Optional) Go to maps.google.com 5) (Optional) Close all opened tabs Actual results: After each of the steps, CPU usage goes up to around 95-99% and stays there for about a minute, while the user (and the site, as far as I can tell) stays inactive. The sites do not play video nor sound nor anything like a game, yet CPU usage is uncommonly high. Screen capture video: https://youtu.be/_Cmw5X29oCk Gecko profile: https://perfht.ml/2Gikmy1 Expected results: I would expect more moderate CPU usage, especially after the page has finished loading and user interaction has stopped. Observed behaviour started with Firefox 57 and continues through all updates since.
Note that my choice of websites to visit is purely illustrative -- CPU usage goes way up into the >95% range no matter what site I'm on, or even when I just have an empty tab (as illustrated in my video). I used fairly static pages with no videos in order to hopefully narrow down the problem more easily. Let me know if there's some other simple/sample site I should visit to bring out the problematic behaviour even more saliently.
Hey Mike, Can you take a look at this profile to see what is going on? Thanks
Flags: needinfo?(mconley)
Hi david, Skimming through the video for a few moments, I haven't yet seen Firefox get close to 95% CPU usage. Certainly, Task Manager is reporting that the sum of all processes is using approximately 95% of your CPU at times, and oftentimes, Firefox is taking up a majority (in the moments I watched, usually in the area of 30-40%), but the rest is a long tail of other processes (System, Trend Micro, OBS, etc). Is there a part of the video in particular I should be looking at where Firefox itself uses upwards of 95% of the CPU? If so, can you give me the timecode?
Flags: needinfo?(mconley) → needinfo?(david.hellfeuer)
Hi Mike, thank you for taking the time to look into this. You are correct, and I probably wasn't clear in my description: While total CPU usage goes up to 90-99%, Firefox itself is only at 30-40% at any given time. So the bits of the video you looked at were representative. The thing is, CPU usage of my system "at rest" and with Firefox closed is generally well below 10%, the surge of general system activity routinely coincides with FF spikes to bring the overall usage to the limit. I cannot tell whether there is a link, whether something FF is doing triggers some OS processes to run amok, but it certainly seems to occur at the same time. I hope this clarifies the report, and I hope it is still relevant enough for you to look into -- even the 30-40% CPU usage seems surprising to me, given the relatively harmless sites (or even empty tabs) and the duration of the spikes. If I can clarify or help in any other way, do let me know. Best, David
Flags: needinfo?(david.hellfeuer)
Hm. Okay, david.hellfeuer - would you mind doing an experiment for me? Can you load those pages again to get the high CPU usage, and then wait about 30 seconds. Do you notice the CPU usage go back down? If not, can you please stop the Profiler, restart the Profiler, wait 10 seconds, and then dump and post a new profile? Restarting the profiler like that will drop the samples from the page load, and give me a better chance to see what's causing the CPU usage _after_ the page has finished loading.
Flags: needinfo?(david.hellfeuer)
Hi Mike, sure, will do! I'll post the results right on Monday (since it's my work laptop, I don't have access to it on weekends). Best, David
Flags: needinfo?(david.hellfeuer)
Flags: needinfo?(david.hellfeuer)
Hi Mike, sorry for the delay, I couldn't run the experiment on my work machine before today. What I found was that with today's nightly (60.0a1 (2018-03-01) (64-bit)) I couldn't reproduce the unusual behavior in the same way. The CPU spikes were lower (~25% when before it was 30-45%), more intermittent (brief bursts rather than staying at peak for a good half minute) and in general shorter (about 20 seconds after requesting a page, CPU usage was down to normal; except for maps.google.com, which still takes a good 30 seconds to normalize). Even the beta that I'm using productively (59.0b12 (64-bit) right now) has gotten better, though not as good as the Nightly yet. From my end, it looks as though the problem is solved. Unless you think there's still something worth looking at? Thanks for your time in any case, David
Flags: needinfo?(david.hellfeuer)
Thanks for getting back, david.hellfeuer. I'll close out this bug as INCOMPLETE then. Thanks!
Status: UNCONFIRMED → RESOLVED
Closed: 7 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.