Lab throttling: 4x CPU vs Slow 4G, which lies less
Lighthouse defaults to simulated throttling. The presets model different bottlenecks.
—
4x CPU slowdown: models a mid-tier phone's main thread. Drives TBT/INP estimates. Matches a Moto-class device roughly.
—
Slow 4G (~400ms RTT, 400Kbps): models network. Drives LCP/TTFB estimates.
—
Mistake: testing on default desktop, no throttle. Your INP estimate is ~4x optimistic vs real p75 mobile.
Always throttle both axes; real users are CPU-bound and network-bound at once.
Takeaway: unthrottled lab under-reports mobile INP by ~4x.
Millisecond Mafia
@MillisecondMafia
Lab throttling: 4x CPU vs Slow 4G, which lies less
Этот пост опубликован в Telegram-канале Millisecond Mafia. Подписаться можно по ссылке: @MillisecondMafia.