You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Problems 1 & 2 are the same thing. Dump analysis shows that all renderers start to consume extra memory. One of them is Brave Extension that is counted in unknown_process and it's the only reason why unknown_process metrics were changed. The consumption of other "unknown" processes was unchanged. Disabling Brave Extension confirms this theory.
Perf run for Chrome shows that they also have a slightly anomaly: 33.3 MiB => 35.8 MiB.
Disabling testing_field_trials (a testing build-in Finch config) for perf run leads to more significant anomaly for Chrome: 24.7 MiB(!) => 30.4 MiB. This behavior is similar to Brave . And yep, disabling testing_field_trials surprisingly saves a lot of memory for Chrome.
And what about production? Comparing beta 102.0.5005.40/103.0.5060.53 (with Finch enabled) show the same thing (see screenshots). The same thing for stable 103.0.5060.114 and for beta 104.0.5112.39.
What exactly start to consume more memory in Chrome? Unfortunately, top level memory dump doesn't provide enough data (all metrics looks unchanged). Chrome chromeperf.appspot.com doesn't have enough precision to point to some commit range. So we need more deep investigation if we want to found a reason. IMHO better to file a ticket and ask chromium folks.
https://brave-perf-dashboard.appspot.com/report?sid=4a087982db1a6c21095c971a6cbc2f8d5e583def592f588a90de6daf44a29d0d
brave/brave-core@v1.41.55...v1.41.56
The tricky thing is that we have a combination of two degradations:
Gpu/browser process memory isn't changed.
The text was updated successfully, but these errors were encountered: