Forum Discussion
Browser and GPU process are much higher compared to Brave
Edge Insiders,
My name is Tim Scudder and I’m a member of the Edge performance team. I wanted to provide an update on this issue: we were able to repro the problem locally, we have a fix coded, validated and are now working to get this change into our next dev drop (estimated to be 76.0.166.0).
We apologize for the inconvenience, but truly appreciate everyone’s help in isolating the problem.
NOTE: We are also aware of a VSync timer tick issue that also has CPU/battery impact that originated upstream in Chromium's codebase. This issue has been fixed upstream and we are also working to make sure the fix is in our next dev drop (again, estimated to be 76.0.166.0)
Regards,
Tim Scudder
"I want to help ensure that new Edge uses resources as sparingly as Edge (Classic)."
Are you for real? What in your view prompted the devs to move to Chromium engine then? Did you look at browser https://www.w3schools.com/browsers/ lately? :)
Opening Edge Classic with a few heavy tabs takes out almost all available memory on an older PC with 4GB RAM, and then it crashes. Chromium in sharp contrast has auto https://developers.google.com/web/updates/2015/09/tab-discarding feature, which can be further enhanced with a choice of extensions, and makes possible to run Chrome on an older and slower PC and devices with numerous tabs open without any user effort. Let me remind you, Windows 10 official https://www.microsoft.com/en-us/windows/Windows-10-specifications requires just 1Ghz CPU with 1-2 GB RAM.
Lets hope, the devs will bring better Edge Classic video playback performance to Edge Preview as promised, while not senselessly discarding numerous features, flags, components, and options offered by Chromium thus limiting use of its wast extensions choice.
"In each case, over a period of 15 +/- minutes of watching Task Manager, an identical CPU usage pattern appeared"
I don't want to call such testing "fake news", but a typical user wouldn't wait 15 min for "identical pattern" to appear to click on the next page link, which would immediately raise Edge Preview CPU usage back to 15-30% for the next 15 min. Again, you're doing great disservice to the team by posting such "optimistic reports", since you can't convince end users whose battery keeps draining twice faster and PC/laptop fans wind-up twice louder after starting Edge Preview compare to other browsers. Lets hope they'll fix it fast.
sambul95 "Are you for real? What in your view prompted the devs to move to Chromium engine then? Did you look at browser https://www.w3schools.com/browsers/ lately? :)"
I've discussed this issue with others in other threads. I don't think that Microsoft's motivations (reducing browser development/maintenance costs for an unpaid front end to paid, profitable business services) are particularly relevant to this thread.
"I don't want to call such testing "fake news", but a typical user wouldn't wait 15 min for "identical pattern" to appear to click on the next page link, which would immediately raise Edge Preview CPU usage back to 15-30% for the next 15 min."
You sound angry, but I don't think that your anger is justified.
In the post you replied to and in a subsequent post, I confirmed the CPU issue on three computers (relatively high end, mid-range and low end, using different processors and graphics), confirmed that the issue is confined to Edge Chromium by installing Google Chrome and testing against that browser, and reported the issue to the Edge team with enough documentation for them to replicate the issue and get started. In other words, I did the basic work needed to report the issues and give Microsoft enough information to understand that the problem is a base-level issue. At this point, it is up to Microsoft to identify the underlying cause(s) and resolve the issue.
I designed the testing I did to eliminate variables, such as Windows 10 visual effects, browser extensions and add-ons, graphically intensive websites, and constantly updating/changing websites. In other words, I stripped the test environment down to the basics, as best I could, and tested four browsers (Edge (Classic), Edge Chromium, Firefox and Chrome), selecting Bing instances to allow the browsers to come to rest, eliminating the effect of ads and other changing environmental variables. If there is a CPU drain at rest, a drain that does not exist in other browsers, then the issue is a base-level issue inherent to the browser. That was what I was trying to check out, and that is what the results showed.
"Again, you're doing great disservice to the team by posting such "optimistic reports", since you can't convince end users whose battery keeps draining twice faster and PC/laptop fans wind-up twice louder after starting Edge Preview compare to other browsers."
It seems to me that I reported objective facts, and that my reports were neither "optimistic" nor "pessimistic". It was not my intent to convince you are anyone else about anything. My intent was to gather the basic facts needed to report the issue to the Edge team as a base issue isolated to Edge Chromium, and that is what I did.
I've done what I planned to do, and all I plan to do. If you think that the testing I did is "fake news", then test as you see fit and report the results of your testing to Microsoft, as I did.
tomscharbach: "We've had a dozen threads requesting that Microsoft do what it can to reduce resource use in Edge Chromium by bundling processes ..."
sambul95: "You might be interested to read this thread: Why are there multiple Chrome instances running even though I only have one window open? Its improving stability, security, and responsiveness."
I initially posted about Chromium's resource-hogging proclivities because I misunderstood the nature of Marco's initial post, which was about excessive CPU usage at rest. If I had understood his initial post, I would have gone right to testing, in order to identify, isolate and document the bug we've reported in this thread.
I'm well aware of the reasons that Chromium uses multiple processes instead of bundling, and if you read the numerous earlier threads in which the topic is mentioned, comments will point you back to 2008 technical documents in which the pros and cons were hotly debated in the early days of Chromium development for the Linux platform.
It seems to me that is a topic you should take to another thread, if you want to do so.
I agree with your observation (stated in your post above this latest edit): "This topic is dedicated to a very important bug." I agree with that, and with your advice to in a more recent thread to "stay focused". Let's not get sidetracked.
Aaron44126 Thanks, Aaron. I was able to replicate your results on a Dell XPS 8920 (i7, 4 physical cores, 8 logical processors). The Edge Chromium Task Manager showed 139.6% CPU use, while at the same moment, Windows Task Manager was showing 20.5% CPU use. I think you've confirmed that the two measure/report CPU usage on differently, and your single/multi core explanation makes sense to me.
It well may be that one CPU core power in https://superuser.com/questions/162590/setting-permanent-process-cpu-core-affinities-in-windows or Chrome https://chromium.googlesource.com/chromium/src/+/master/docs/threading_and_tasks.md is enough per Chromium process to render most webpages. It looks like Browser Task Manager is designed to reflect that as you suggested earlier. :)
sambul95 I ran the SunSpider JavaScript benchmark and here you can see an Edge process using more than 100% CPU. (At one point it spiked to over 200% but I missed grabbing a screen shot for that.)
"run with admin privileges"
You can start a new topic about Edge Chromium Roadmap, sure many folks will contribute. This topic is dedicated to a very important bug. Edge Chromium Program Manager was working for Google Chrome Security team just months ago, so you can address your concerns to the right guy if they are justified. :)
"We've had a dozen threads requesting that Microsoft do what it can to reduce resource use in Edge Chromium by bundling processes"
You might be interested to read this thread: https://superuser.com/questions/461552/why-are-there-multiple-chrome-instances-running-even-though-i-only-have-one-wind Its improving https://docs.microsoft.com/en-us/windows/desktop/ProcThread/about-processes-and-threads, https://www.chromium.org/Home/chromium-security/site-isolation, and https://blog.chromium.org/2008/09/multi-process-architecture.html. While I agree OS developer company is better positioned to cut on browser resource usage, if its various teams interact properly. For that to happen, a top level MS exec must be keenly interested in Edge success, given current minuscule market share and revenue from it. Now it looks more like a prestige project, though they put 200 heads on it. Keep pushing on the right buttons. :) Or https://www.ghacks.net/2015/02/08/save-memory-in-chrome-by-using-one-process-per-site/ on your own.
sambul95 "My main concern is, Edge Preview devs have an approved roadmap, which directs them to concentrate on full set feature transfer, and moving from Chrome to MS services. Which means, performance optimization issues may be very last ones on their mind right now."
Your comment suggests that you have access to the Edge team's official roadmap in some form. I understand that you may have to keep the details close to your chest, but could you comment about whether or not removing the Edge Chromium's ability to run with administrator privileges is on the roadmap?
I don't run Chrome on my computers, in part, because I have five or six concerns about Chrome's security. The ability to run with administrator privileges is high on that list of security deficiencies, because it isn't all that hard to bypass/escape a sandbox.
Following the same logic, can you give some alternative links for Chromium Task Manager? I'm merely asking for sources, not sure how it measures the load. :)
It's pretty common knowledge that the CPU usage percent in Windows Task Manager is normalized across the CPU's logical cores. A quick search would have told you. :)
For example:
https://www.quora.com/What-is-the-meaning-of-CPU-usage-in-a-windows-task-manager
https://superuser.com/questions/994191/what-does-cpu-column-means-on-process-tab-on-task-manager
tomscharbach: "I was wondering why the metrics were so different. Now I know. "
sambul95: "I'd ask for some references before saying that."
I accepted Aaron44126's explanation (different ways of dealing with cores) because it made sense.
I know one half of the equation: Windows 10 Task Manager tracks usage across all logical processors (typically two per physical core) and measures % of CPU cumulatively across the logical processors:
I don't use Google Chrome on Windows or Chromium on Linux, so I'm not familiar with the Chromium-based task manager. Accordingly, I don't know anything about the other half of the equation -- how the Chromium-based browser Task Manager tracks usages across physical/logical cores/processors.
You seem to have a high level of technical expertise, and you seem to be challenging Aaron44126's explanation, so let me ask you this: To what do you attribute the differences in CPU usage measurement between Windows 10's Task Manager and Edge Chromium's Task Manager when measuring CPU usage under identical conditions? I'm curious to know why you think that this is happening, if you don't accept Aaron44126's explanation. Clearly the two task managers are measuring something differently.
I too stopped using Chromium Edge too due to performance problem. I switched back to Google Chrome and was surprised that my laptop got quiet again :-(. Very sad. I hope MS will fix this performance issue very soon.
I reported this issue via feedback and attached screenshot with task manager showing CPU and GPU consumption in idle state when only start page is open.
I'm sorry to repeat my basic question: where did you get this info from? :) Can you support it by any web links or screenshots?
sambul95 I stated nothing of the sort. I just stated that what the "CPU %" shown in Task Manager means is different between the two... task managers. To compare the two you have to multiply or divide by the number of logical cores in your system. If one process within Edge process uses more than one CPU core it will register higher than 100% in Edge task manager
"What processor are you guys using to get CPU use results in the 20-30% range, if CPU power is the driver on this issue?"
It was already answered in this thread. Look no further than Win 10 official specifications: 1Ghz CPU, 1-2 GB RAM. That MUST be Edge devs target as per MS Policies and common sense, since Edge is internal Windows app. Don't forget, near same code is likely used not only on PCs, but Windows Mobile devices of various gens, Tablets and very basic and cheap Atom and such student notebooks sold in huge quantities.
"I was wondering why the metrics were so different. Now I know. "
I'd ask for some references before saying that.
Are you stating that Edge Chromium uses only ONE core of ANY CPU? Where did you get it from? Also, I don't need to look at any Task Manager - when my laptop fans are sounding like turbines all the time I know Edge is idling, now uninstalled. :)
My main concern is, Edge Preview devs have an approved roadmap, which directs them to concentrate on full set feature transfer, and moving from Chrome to MS services. Which means, performance optimization issues may be very last ones on their mind right now. So no such bugs might be fixed in forceable future, thus cutting off current enthusiastic but fast shrinking testers pull. There always be some hardcore folks, knowing little about tech, but asking to add this or that classic button or check box thinking Edge team "forget" it, so targeted "web noise level" will be maintained.
Aaron44126 "Keep in mind: It appears that there is a discrepancy between the number reported between Windows Task Manager and the browser's Task Manager. Windows Task Manager reports the CPU use where 100% would mean that all logical CPU cores are fully loaded. (i.e. In a system with 8 logical cores, 100% would mean that all 8 are under a full load whereas a single-core load would be 12.5%.) The browser task manager reports 100% for one core worth of full load, so an 8-core load would be 800%. Something to keep in mind when making comparisons."
Thanks, Aaron. I was wondering why the metrics were so different. Now I know.
Keep in mind: It appears that there is a discrepancy between the number reported between Windows Task Manager and the browser's Task Manager. Windows Task Manager reports the CPU use where 100% would mean that all logical CPU cores are fully loaded. (i.e. In a system with 8 logical cores, 100% would mean that all 8 are under a full load whereas a single-core load would be 12.5%.) The browser task manager reports 100% for one core worth of full load, so an 8-core load would be 800%. Something to keep in mind when making comparisons.
Marco Mollace ""Edge Chromium isn't using more than 6-7% at rest." CVD, there's an issue. It should stay around 0%"
Absolutely. That's what I've been documenting throughout this thread, as carefully as possible, using Windows Task Manager, comparing browsers. Edge Chromium uses CPU resources at rest; other browsers (Edge (Classic), Firefox, Chrome) don't. Something is not right with Edge Chromium, or it would not be using CPU at rest.
On a related topic, I did a comparison using the in-built Browser Task Manager in Chrome and Edge Chromium. The difference in CPU use shows up there, as well, and the numbers are closer to the results you and others have been reporting.
This is the side-by-side results for Edge Chromium (left) and Google Chrome (right) on the Dell Inspiron 3185:
- "Edge Chromium isn't using more than 6-7% at rest."
CVD, there's an issue. It should stay around 0% Marco Mollace "Your PCs are too powerful to notice the high CPU usage of Edge C."
Okay. I'm curious now.
I do my Edge Chromium testing on three computers -- one high end, one mid-range, and one low end. The low end computer (Dell Inspiron 3185, A6 9420 CPU, 4gb RAM, AMD R5 integrated graphics) is an entry-level computer selling for $175-$200. I guess that it is possible to go lower than that (say, a Celeron 3060) but now much lower. Even on the Inspiron 3185 (as reported above), Edge Chromium isn't using more than 6-7% at rest.
What processor are you guys using to get CPU use results in the 20-30% range, if CPU power is the driver on this issue?
tomscharbach Your PCs are too powerful to notice the high CPU usage of Edge C.
Also, like title of this report states, I'm talking about internal browser task manager, which clearly show how buggy is current version of Canary and Dev build, compared to other Chromium browser like Chrome, Brave etc.
It's clearly an issue of Edge C. If you want definitive proof, search online leaked builds of Edge C and compare.
If you got 1% of CPU usage, we got 20%.
P.S: Guys, take it easy.
sambul95 "You don't have to go that far. Just open any INTERNAL Edge webpage - just one page, not plenty irrelevant Bing tabs, and you'll see the same high CPU load, as shown on my screenshots above. What can be more objective than showing a browser crippling under no load at all and without any tabs open? :)"
I cannot replicate your results. I opened the "Settings" page on Edge Chromium and Google Chrome on each of my three computers. The "Settings" tab, in each case, was the only tab open. The browser, in each case, was set to half screen and Task Manager was open in the other screen, so that I could check CPU use.
These are the results I got:
On the Dell XPS 8920:
EDGE CHROMIUM
Version 76.0.159.0 (Official build) dev (64-bit)
Tab: Settings
No Extensions, Add-ons or Modifications
CPU USE: 1.8%
Dell XPS 8920
CPU: Intel i7
RAM: 16gb
GPU: AMD Radeon 580
Windows 10 1809 Build 17763.503
Visual Effects (Transparencies, Shadows etc) Disabled
GOOGLE CHROME
Version 74.0.3729.157 (Official Build) (64-bit)
Tab: Settings
No Extensions, Add-ons or Modifications
CPU USE: 0.1%
Dell XPS 8920
CPU: Intel i7
RAM: 16gb
GPU: AMD Radeon 580
Windows 10 1809 Build 17763.503
Visual Effects (Transparencies, Shadows etc) DisabledOn the Dell Latitude 7280:
EDGE CHROMIUM
Version 76.0.159.0 (Official build) dev (64-bit)
Tab: Settings
No Extensions, Add-ons or Modifications
CPU USE: 4.6%
Dell Latitude 7280
CPU: Intel i5
RAM: 8gb
GPU: Intel 620
Windows 10 1809 Build 17763.503
Visual Effects (Transparencies, Shadows etc) Disabled
GOOGLE CHROME
Version 74.0.3729.157 (Official Build) (64-bit)
Tab: Settings
No Extensions, Add-ons or Modifications
CPU USE: 0.0%
Dell Latitude 7280
CPU: Intel i5
RAM: 8gb
GPU: Intel 620
Windows 10 1809 Build 17763.503
Visual Effects (Transparencies, Shadows etc) DisabledOn the Dell Inspiron 3185:
EDGE CHROMIUM
Version 76.0.159.0 (Official build) dev (64-bit)
Tab: Settings
No Extensions, Add-ons or Modifications
CPU USE: 5.9%
Dell Inspiron 3185
CPU: AMD 9420e
RAM: 4gb
GPU: AMD R5
Windows 10 1809 Build 17763.503
Visual Effects (Transparencies, Shadows etc) Disabled
GOOGLE CHROME
Version 74.0.3729.157 (Official Build) (64-bit)
Tab: Settings
No Extensions, Add-ons or Modifications
CPU USE: 0.0%
Dell Inspiron 3185
CPU: AMD 9420e
RAM: 4gb
GPU: AMD R5
Windows 10 1809 Build 17763.503
Visual Effects (Transparencies, Shadows etc) DisabledI am not doubting your results or the validity of the issue you and others in this thread are raising. I am simply saying that I cannot replicate your results.
I know, from the testing I've done (see several comments throughout the thread) that Edge Chromium has a CPU use problem, because Edge Chromium consistently and persistently uses CPU resources at rest, while other tested browsers (Edge (Classic), Firefox, Chrome) do not. I've reported that issue, pointing to this thread.
I also know that Edge Chromium's CPU use at rest manifests in indirect relationship to CPU power (that is, the more powerful the CPU, the lower the percentage of CPU use; the less powerful the CPU, the higher the percentage of CPU use). That is expected behavior.
I know, in addition, that CPU use jumps very high under load. I have been using the browsers at half screen, with Task Manager running in the other half, and I routinely see CPU use jump as high as 75-85% when loading an ad-intensive web page. That seems to be unrelated to the browser used (that is, I get the same or similar results using Edge Chromium, Edge (Classic), Firefox and Chrome), although I don't have the testing platforms sufficient to cross-test to document my impression. All I can say on that score is that all the tested browsers use a lot of CPU temporarily while loading ad-intensive pages.
At any rate, I have done what I can at this point. I intend to remove Google Chrome from my computers and put the issue in Microsoft's hands.
I know that this is very frustrating for you and others. Something is clearly causing the issue you describe. I just can't replicate, identify or isolate the issue beyond the results I've reported. I hope that Microsoft will get to the bottom of it quickly. I also hope that Microsoft will get to the bottom of the "CPU use at rest" issue that I've identified and reported. The two issues may well be related.
The screenshot below shows the test results for the XPS 8920, just for the record:
You don't have to go that far. Just open any INTERNAL Edge webpage - just one page, not plenty irrelevant Bing tabs, and you'll see the same high CPU load, as shown on my screenshots above. What can be more objective than showing a browser crippling under no load at all and without any tabs open? :)
""Fake" part in your "news" seems the limit CPU load difference to negligible 1.5% at idle, where your definition of "idle" is shifted to 15 min wait, while real difference is 15-30% on an average PC after 2 min wait."
The CPU usage difference appeared almost immediately. I ran the test for 15 minutes to make certain that the issue was persistent. I ran the test on three computers to make certain that the issue was consistent. I ran the test when the four browsers were at rest to make certain that the issue was in the browser itself.
"Meanwhile, what makes you think your report is the only one submitted to devs by now? :)"
I don't, and I've not said or suggested that mine was the only report submitted.
If the issue is as crippling under load as you and others report, I would hope that there would have been many reports submitted, a healthy fraction of which would have been systematic, objective comparison testing under load.
"Fake" part in your "news" seems to limit CPU load difference to negligible 1.5% at idle, where your definition of "idle" after a webpage load is shifted to 15 min wait, while real difference is 15-30% on an average PC after 2 min wait. Meanwhile, what makes you think your report is the only one submitted to devs by now? :)
I'm optimistic about devs fixing this bug or "feature" sooner or later, in fact more optimistic than for the prospects of adding smiles and picture/video links to this heavy JS fitted board with almost no text area format controls. :)
The only thing I'd suggest Edge team to do to raise the browser popularity is ensure streaming video playback like torrents, including in a separate window via NACL, since it would allow to capitalize on Edge better integration with OS in video playback, and may attract plenty of weaker hardware users dreaming of HD torrents smooth direct playback without prior full download. This is REAL https://blogs.windows.com/windowsexperience/2016/07/13/get-better-quality-video-with-microsoft-edge/#Kdcsbg1KWl21MxKF.97 Edge advantage, I doubt anything else would change current browser usage statistics. Meanwhile, using separate process per tab is required for stability, and some tabs open several processes for faster parallel calcs.
