TTI Metric in PWA & performance reports do not match

[mdiblasio]

Provide the steps to reproduce

1. Run LH on https://swipelife.tinder.com/

What is the current behavior?

TTI in PWA report (11s) and performance report (5s) do not match, making it unclear if the metric is measured different in each report (e.g. is throttling different? Does TTI in performance not take into account network requests?)

What is the expected behavior?

TTI should match

Environment Information

• Affected Channels:
• Lighthouse version: 3.2.0
• Node.js version:
• Operating System:

[patrickhulce]

Thanks for filing! How did you run lighthouse @mdiblasio? The PWA audit is making sure your TTI is fast enough on mobile networks so if you didn't run with one of our approved presets the numbers will be different.

[exterkamp]

We might want a visual indication between the two TTI's if they are derived with different methods. I know I was confused by this since they are different. Something like this on the audit's TTI explanation:
Interactive (on Simulated Mobile Network) at 11s vs just Interactive at 11s

[cheneytsai]

This was run through WPT (Moto G4, Chrome, 3G, Dulles) which under my understanding should run LH with the same conditions which in this case is a 3G network. This results in the following metrics:

First Interactive (WPT): >5.9s
Time to Interactive (LH - Performance): 5.0s
Interactive on 3G (LH - PWA): 11s

Webpagetest Link

[patrickhulce]

Thanks for sharing the report @cheneytsai!

#6162 and #6286 should help with some of the problems here. Locally, simulated performance reflect WPT results quite accurately, but in the load-fast-enough-for-pwa we're double throttling the CPU which is probably what's inflating the numbers. I'll do a followup to #6286 to fix this.