Run the following experiments on Beta. 

1- De-prioritize hidden frame tasks to low priority. 
2- De-prioritize subframe tasks to low priority during loading.
3- De-prioritize throttleable tasks to low priority during loading. 
4- De-prioritize subframe throttleable tasks to low priority.
5- De-prioritize ad frames to low priority. 
6- De-prioritize ad frames to low priority during loading.
7- De-prioritize background pages to low priority.  

All of these experiments were run on Canary/Dev. Experiments 1,2,3 seem to have a good impact on our loading metrics for low memory devices. Here are the results for the first four experiments on Canary/Dev:

https://uma.googleplex.com/p/chrome/variations/?sid=d1e5079dae6b66abfa1892f4a85a991c

The last experiment isn't have any good impact, and we are interested in running it on Beta as well to get a better understanding of why is this happening.