AMP cache versus Cloudflare

I was curious what AMP would do to the performance of an average blog post on my site, so I ran a series of tests with AMP for WordPress (0.6.2) versus just a cached post served via Cloudflare.

As a sample post, I picked a fairly long blog post with a decent amount of images (including some animated GIFs).

As for caching and image optimisation on this site, I rely on WP Rocket and Imagify (obviously), as well as a lean theme from foxland.fi that’s crazy fast out of the fox … box. (I still made a child theme and tinkered around.)

I used GTmetrix, Lighthouse in Chrome, and Google PageSpeed Insights on each of these page views:

• The AMP page of the article, served from my own domain.
• The same AMP page, served from AMP cache.
• The single post without AMP, served via Cloudflare.

Update: Forgot to save my GTmetrix reports, so they’re gone. Meanwhile, Tim Kadlec has explored how fast AMP really is.

GTmetrix

Tests from Vancouver/Canada, on Chrome/Android, Galaxy Nexus

Lighthouse

Tests from Lighthouse audit panel in Chrome

Google PageSpeed Insights

PSI doesn’t provide permalink archives, so screenshots will have to do.

Conclusion

AMP for WordPress 0.6.2 supports only “paired mode” (something that will change in 0.7), so we’re talking about a visual appearance that varies greatly between AMP and non-AMP mode.

The gain of 500 ms in loading time achieved through serving the page via AMP cache instead of plain Cloudflare seems quite significant and would definitely make me consider AMP for the future if my site had a commercial background.

For now, I have deactivated AMP again, as I continue to weigh its approach against its impact on what we call an open web.