All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=844445

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=f9760d0c9b83e880d5b90393c19847ae8aa8996b0d4a697770aea6cc5ee476be


Bot(s) for this bug's original alert(s):

chromium-rel-win10
win-high-dpi

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12c07054240000

📍 Found significant differences after each of 3 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/12c07054240000

Make --site-per-process the default on ToT via fieldtrial_testing_config by lukasza@chromium.org
https://chromium.googlesource.com/chromium/src/+/fb1ccf02ee8ca79e1404abfd3a3a7d540b7d2dbd

Disable flaky SystemHealthBenchmarkSmokeTest test by jonross@chromium.org
https://chromium.googlesource.com/chromium/src/+/ddf61b968ad5a0225a9efcf3358dbed0a160a341

Add latency_info to original event list by eirage@chromium.org
https://chromium.googlesource.com/chromium/src/+/4ef6344563acf9961f28f13e36eaebdffd5077e8

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

AFAICT the graphs in #c1 and the pinpoint job in #c2 and #c3 are measuring memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg.  It is expected that site-per-process uses more memory - on average we expect to increase memory usage (e.g. as measured by Memory.Total.PrivateMemoryFootprint) by around 10%.

The regression in these graphs is 25% (in the second one, the big jump points to  issue 835840 ). Any idea why there's such a large gap between the regression here vs the expected 10% from PrivateMemoryFootprint?

The benchmarks usually execute navigations to various pages in the same tab whereas real users on desktop tend to have more than one tab. Once there is more than one tab there is likelihood processes can be reused for iframes common to the various tabs. So what we test with benchmarks for desktop is not what the usage pattern is in reality, which I think is the root cause of the mismatch.