Which graph, Petr?

No idea. Looks like a perf dashboard failure...

Here are the graphs: https://chromeperf.appspot.com/group_report?bug_id=628265

Started bisect job https://chromeperf.appspot.com/buildbucket_job_status/9005151437909870192

===== BISECT JOB RESULTS =====
Status: completed


===== SUSPECTED CL(s) =====
Subject : Add benchmark that imitates news reading.
Author  : ulan
Commit description:
  
Each story in the benchmark does the following:
1. Load the main page.
2. Open and scroll the first news item.
3. Go back to the main page and scroll it.
4. Open and scroll the second news item.
5. Go back to the main page and scroll it.
6. etc.

This workflow is interesting because it is close to real world
and generates a lot of garbage by creating and discarding pages.

A couple of websites crash with OOM on svelte device, so by
tracking and optimizing for this benchmark we can improve
user experience on low-memory devices.

*** NOTE TO SHERRIF ***
Regressions are expected in system_health.memory_mobile and
system_health.memory_desktop for the following stories:
- load:news:cnn
- load:social:facebook
- load:news:qq
- load:news:reddit
- load:social:twitter

If there are failing tests in form of benchmarks.system_health_smoke_test.SystemHealthBenchmarkSmokeTest.*
please do not revert this CL but instead disabling failing
test by adding them into the _DISABLE_TESTS list in
tools/perf/benchmarks/system_health_smoke_test.py

BUG=589726
CQ_INCLUDE_TRYBOTS=tryserver.chromium.perf:android_s5_perf_cq;tryserver.chromium.perf:mac_retina_perf_cq

Review-Url: https://codereview.chromium.org/2118293002
Cr-Commit-Position: refs/heads/master@{#405212}
Commit  : 5b1847df818e5be4592e9a6ac4832350bb663023
Date    : Wed Jul 13 17:39:28 2016


===== TESTED REVISIONS =====
Revision         Mean     Std Dev  N  Good?
chromium@405107  143846   1624.59  5  good
chromium@405164  148019   1021.52  5  good
chromium@405191  143525   1811.76  5  good
chromium@405209  144469   817.384  5  good
chromium@405211  145614   1045.37  5  good
chromium@405212  6171096  929039   5  bad    <--
chromium@405213  6586858  1197.31  5  bad
chromium@405216  4925398  929430   5  bad
chromium@405221  6170636  928779   5  bad

Bisect job ran on: win_8_perf_bisect
Bug ID: 628265

Test Command: src/tools/perf/run_benchmark -v --browser=release --output-format=chartjson --upload-results --also-run-disabled-tests system_health.memory_desktop
Test Metric: load_social-memory:chrome:all_processes:reported_by_chrome:skia:effective_size_avg/load_social_facebook
Relative Change: 4189.74%
Score: 99.9

Buildbot stdio: http://build.chromium.org/p/tryserver.chromium.perf/builders/win_8_perf_bisect/builds/2097
Job details: https://chromeperf.appspot.com/buildbucket_job_status/9005151437909870192


Not what you expected? We'll investigate and get back to you!
  https://chromeperf.appspot.com/bad_bisect?try_job_id=5907566501036032

| O O | Visit http://www.chromium.org/developers/speed-infra/perf-bug-faq
|  X  | for more information addressing perf regression bugs. For feedback,
| / \ | file a bug with component Tests>AutoBisect.  Thank you!