Could you please provide a link to the graph showing the growth? This is all I can find:
https://uma.googleplex.com/p/chrome/histograms/?endDate=20181124&dayCount=7&histograms=Memory.Gpu.PrivateMemoryFootprint&fixupData=true&showMax=true&filters=platform%2Ceq%2CM%2Cchannel%2Ceq%2C4%2Cisofficial%2Ceq%2CTrue&implicitFilters=isofficial

Thanks.

+couple more GPU folks to have more eyes on this.

https://uma.googleplex.com/p/chrome/timeline_v2/?sid=c7280c6c32e4dc7c0ed83f75a3ada8d3 is the graph I'm looking at.

It's curious that while gpu has increased, the browser, renderer, and total memory have all decreased: https://uma.googleplex.com/p/chrome/timeline_v2/?sid=de8eeedca06d16058d7a0a71401cf52c

I also checked out oop-raster (which was in finch on canary briefly) which moved memory from renderer to gpu, but that appears to make things better overall: https://uma.googleplex.com/p/chrome/timeline_v2/?sid=270af84f43d38832993dc402c753a6af

Is it possible to view this data normalized by the number of renderer processes? Could it be the case that users are on average leaving more tabs open, since the GPU process services all renderers?

I don't think you can do that, but Tabs.TabCount and Tabs.TabCountPerWindow are both flat on Mac: <https://uma.googleplex.com/p/chrome/timeline_v2/?sid=8b4c15c98946408923e279411636ea34> so I don't think the data we do have supports that hypothesis.

Thanks for the data. Guess we need another hypothesis.

For what it's worth, this metric is still climbing: the median is now at 165MB, up from 149MB a month ago and up from 159MB when I filed this.

This appears to have gone back down to a 147MB median over the last month.