Forum Discussion
Browser and GPU process are much higher compared to Brave
- May 22, 2019
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 https://bugs.chromium.org/p/chromium/issues/detail?id=962784 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
This bug has just hit dev build as well, draining battery like crazy.
I did some comparison testing on CPU use this morning, opening instances of Edge (Classic), Edge Chromium and Firefox Quantum, each with 6 Bing tabs open, all in static state, on each of three Windows (a Dell XPS 8920, Intel i7, 16gb RAM, AMD Radeon 580 graphics; a Dell Latitude 7280, Intel i5, 8gb RAM, Intel 620 graphics; and a Dell Inspiron 3185, AMD a9420e, 4gb RAM, Radeon R5 graphics). I run Windows without visual effects (transparencies, shadows and so on) and I run all three browsers in native state, so no extensions or add-ons are running that could complicate results. I am using the most recent Dev build (76.0.159.0).
After the browsers were opened an had a minute or two to settle down to static state, I opened Task Manager on each of the three computers and observed CPU use of the three browsers on each computer.
In each case, over a period of 15 +/- minutes of watching Task Manager, an identical pattern appeared that was both persistent on each computer and consistent between the computers: Edge Chromium used 1-1.5% of CPU resources when in the static state described, but neither Edge (Classic) nor Firefox Quantum used CPU resources in the static state described. This snapshot from the Dell XPS 8920 illustrates what I observed:
A caveat: I checked several similar Task Manager screenshots (comparing Edge (Classic) and Edge Chromium, but not Firefox) that I used in a discussion related to Chromium-based process handling. At the time I was using an earlier Dev build (I don't recall which, but I think that it was the first Dev release). I saw similar results in snapshot form (that is, Edge Chromium using 1-1.2% of CPU at rest, Edge (Classic) at 0%). The screenshots from that time suggest that the CPU use pattern is not new with the most recent builds.
I don't know whether this pattern is related to the issue reported by Marco Mollace with respect to Brave or not, and I am not in a position to compare CPU use between Edge Chromium and Chrome, which would seem to me to be a more definitive and useful test.
All I can suggest is that you run similar comparison tests between Edge Chromium and Chrome, and see if the results are similar or dissimilar. If Edge Chromium is using CPU resources in static state, but Chrome is not, then Edge Chromium has an issue that is not present in Chrome, and that should be reported as a bug.
- tomscharbachMay 18, 2019Bronze Contributor
I installed Chrome on the XPS 8920 and ran the CPU comparison again, this time comparing CPU use on Edge (Classic) Windows 10 Version 17763.503, Edge Chromium Version 76.0.159.0, Google Chrome Version 74.0.3729.157, and Firefox 66.0.5, with identical conditions to the previous test (6 Bing instances open on each browser, all Bing instances idle, stock builds of each browser without extensions or add-ons, running on Windows 10 without visual effects.
The results confirmed to me that the CPU use at idle issue is confined to Edge Chromium. Google Chrome, like Edge (Classic) and Firefox, does not use CPU resources at idle. Edge Chromium, and Edge Chromium alone uses CPU resources (typically in the range of 1-1.5%, sometimes higher) at idle.
I've reported the bug, pointing to this discussion thread,