All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=883619

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=267a86f862cfd2b63f151f46854f043ddc6fb205b2ef832d3203d2851effbb7d


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

cros-elm

https://crosbolt.teams.x20web.corp.google.com/prod/crosbolt/time_series/graphics_WebGLAquarium/avg_fps_1000_fishes.html?num_versions=-1&sku_filter=mtk

The graphs for minnie are very clean
https://chromeperf.appspot.com/report?sid=ec0bc1a05a357e93b9655d72aa912ffedc0cf54d8d16232d033095ffcef23289

With a stock 11056.0.0/71.0.3544.0 image I get the same numbers (regression) as on the graph:
  avg_fps_0050_fishes    40.9928796189
  avg_fps_1000_fishes    32.718679368

But when I build Chrome at 71.0.3553.0 (#591288) the regression is gone.

So maybe we just need to wait for an uprev.

That said the regression is also on the branch.

And indeed on #589076 I get 40.4fps and 32.5fps, so the regression got fixed on tot.

I have kicked off a bisecting for this issue. While it's running, let me share some information might be related.

Other serious graphics regressions at the same range:
 - issue 879985
 - issue 879962 (the same culprit CL as 879985)
   This one got a fixed at #590230 and is requesting to merge to M70

Probably the latter #590230, as my bisect ended somewhere in that range before I went home (still had an 80 change gap).

Via bisect, I found the culprit CL is
https://chromium-review.googlesource.com/c/chromium/src/+/1188848
the same as other regressions mentioned in comment 6.

So, we just need to wait for uprev and merge.

I can confirm that on minnie this is fixed by #590230.