GoogleChrome
Update the mobile preset
Feature request summary Can the mobile preset speed be increased, it's an issue that has been raised and dismissed before
What is the motivation or use case for changing this? The mobile landscape has changed dramatically over the last 3 years. The 85th percetile has to be higher than 1.6mbps
https://infrequently.org/2024/01/performance-inequality-gap-2024/#test-device-recommendations now suggests a Samsung Galaxy A51 or Google Pixel 4a as the benchmark, and a P75 estimate of 7.2Mbps down, 1.4Mbps up, and 94ms RTT.
How is this beneficial to Lighthouse? It will bring the audit more in line with real world averages
Apologies for the lengthy comment but I’m new to this project and wanted to note support for @talktomedia-rob’s concerns and add a note that seems quite related and also in need of addressing.
Regular updates needed
I agree. The network throttling and device emulation need to be updated regularly to keep up with the landscape. Basing this on actual data (as the CWVs are) makes the most sense. LCPs seem to be 2-8x higher (example showing 1.9s LCP in CWV vs 15.8s LCP from the test) in the diagnostics vs the CWV’s 75th percentile. If that’s intentional, it’s confusing, and it’d be helpful to understand the reasoning behind it.
Reproducibility needed
But even if we are stuck at 1.6 mb/s (1638 kb/s), this would be much more manageable if it could be reproduced outside of Lighthouse’s simulated throttling, which can only be shown as an unthrottled trace.
Unthrottled trace compares poorly
So we’ll get an LCP of 4.5s in the Lighthouse tab:
But in the unthrottled trace…
we’ll get an LCP of just 0.19s:
Performance tab throttling doesn’t help, either
In the Performance tab, when using the CPU 4x slowdown (to match Lighthouse) and an exact match of Lighthouse’s reported network throttling values…
The 4.5s LCP from the Lighthouse tab is only 0.29s:
This proves that something is dramatically different, in this particular case by 15x, from what can be reproduced by using the same values with devtools throttling. So I’d argue that fixing this discrepancy is even more important (in terms of providing useful tools to help developers improve their site performance) than increasing the throttling speed and emulated device performance that Lighthouse uses.
