All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=919847

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=389324dbceb009d2b23a9b2e62c2f4df2fce61124eb917e571d154dba42a043a


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

webrtc-linux-large-tests
webrtc-win-large-tests

The changes to vp9 are expected, but not the changes to vp8. There seems to be increase in memory usage for some of the vp8 screenshare and simulcast tests. I don't see anything obvious from the libvpx roll.

Erik do you know of any change in webrtc that would cause the vp8 change? 

I  don't see any relevant webrtc changes close to this regression. It's changes to the build system  and removal of dead code....

Was not able to reproduce the memory increase for vp8 on my linux machine, and don't see anything in the roll that would cause this. The reported regressions for vp8 memory increase are only on windows and not so large (1-3%), so will close this for now.