All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=825518

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=c8775241e563652864e198e805db20fe2f719e409146b34e6aa6f897f83e0d61


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

chromium-rel-mac11-air
chromium-rel-mac11-pro

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12e7f43d440000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/12e7f43d440000

Switch back to uploading GpuMemoryBuffers in VideoRendererImpl. by dalecurtis@chromium.org
https://chromium.googlesource.com/chromium/src/+/5c175997b4d9c42c7c3ad40468e7b5ab9950ca69

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

This is a revert to the old path, all metrics are the same as prior to the decoder based GMB approach. The interesting one is:

https://chromeperf.appspot.com/report?sid=6365d6948f0fefecfc49c2d31dad358eb6a8bedb1362f0b4d7065c32b165a249&rev=544617

~3W (!) improvement with decoder based approach. I think this is likely because it's using 100mb more of memory due to bursts of outputs which allow the system to sleep more between outputs.