Ancestors

Toot

Written by Zach Leatherman :11ty: on 2025-01-22 at 15:37

A few Eleventy build server image optimization tests for the project I’m currently working on (×1088 images):

formats: webp, auto =>

❄️ cold cache: 4m 33s

🔥 warm cache: 36s

formats: avif, webp, auto =>

❄️ cold cache: 11m 32s

🔥 warm cache: 38s

Do y’all think the AVIF build cost is worth it if the build server cache is ~always warm?

=> More informations about this toot | More toots from zachleat@zachleat.com

Descendants

Written by Kornel on 2025-01-22 at 15:54

@zachleat Yeah. You should be getting better quality, sharper color.

Number of browsers that support WebP but not AVIF is shrinking, so maybe save some time by dropping the WebP alternative?

=> More informations about this toot | More toots from kornel@mastodon.social

Written by Zach Leatherman :11ty: on 2025-01-22 at 16:15

@kornel valuable advice, thank you!

=> More informations about this toot | More toots from zachleat@zachleat.com

Written by Michelle Barker on 2025-01-22 at 16:23

@kornel @zachleat In case it affects your opinion, @fershad wrote about the relative power consumption of different image formats a while back: https://fershad.com/writing/power-consumption-jpeg-webp-and-avif/. Based on those findings I wouldn’t want to drop WebP. Things might’ve changed since then though.

=> More informations about this toot | More toots from michelle@front-end.social

Written by Eric Portis on 2025-01-22 at 16:42

@michelle @kornel @zachleat @fershad I discussed this article with @jonsneyers when it was published. He pointed out that the energy used for network transfer from the server to the client should be accounted for; especially on mobile (radio), this costs much more than CPU decoding. He raised two other points:

• Energy used to light up the display > both network and decode costs

• Encode costs, esp. for AVIF, dwarf decode costs. Worth thinking about # of loads.

=> More informations about this toot | More toots from eeeps@front-end.social

Written by Kornel on 2025-01-22 at 16:55

@michelle @zachleat @fershad I don't think this should affect anyone's decision about use on regular web pages. The energy usage in absolute terms is too tiny to matter, even multiplied by millions it's still small.

But from overall energy efficiency perspective you should also consider that larger files will use relatively more energy on all the network equipment that sends them.

Computers and displays also use energy while waiting for a page to load. Faster pages reduce wasted time.

=> More informations about this toot | More toots from kornel@mastodon.social

Written by Zach Leatherman :11ty: on 2025-01-23 at 22:33

@kornel @michelle @fershad fwiw I went with your advice and I’m using avif, auto for now!

=> More informations about this toot | More toots from zachleat@zachleat.com

Written by fershad on 2025-01-24 at 12:36

@kornel @michelle @zachleat @eeeps thanks for this conversation. I understand the comments about the network energy, especially for radio networks, I didn't test that segment because:

1. There's no reliable means to measure it

2. Network energy is (as far as I know) pretty constant https://fershad.com/writing/website-carbon-beyond-data-transfer/#data-transfer-network-energy-usage

Eric's points on encoding are very valid, as is Kornel's point about efficiency. Both should be measured further in a more end-to-end kinda test scenario.

=> More informations about this toot | More toots from fershad@indieweb.social

Written by elle mundy on 2025-01-22 at 15:56

@zachleat 2 seconds for a more compatible, higher efficiency codec? yeah i’d say so. that time difference may even be just noise depending on what else the build server is doing

=> More informations about this toot | More toots from exchgr@mastodon.world

Written by Zach Leatherman :11ty: on 2025-01-22 at 16:19

@exchgr yeah 100% I think the uneasiness there is mostly around the the reliability of the build cache — largely out of my control

=> More informations about this toot | More toots from zachleat@zachleat.com

Written by elle mundy on 2025-01-22 at 16:21

@zachleat oo yeah that might be concerning then. not sure if this helps or what the use case is, but i’ve been using cloudflare images, which handles the transcoding for me (although it’s limited in some slightly frustrating ways)

=> More informations about this toot | More toots from exchgr@mastodon.world

Written by Snugug on 2025-01-22 at 15:59

@zachleat absolutely. AVIF's widespread support is why I need to jump through hoops to support HDR images, and find it generally better than webp

=> More informations about this toot | More toots from snugug@mas.to

Written by Zach Leatherman :11ty: on 2025-01-22 at 16:19

@snugug thank you!

=> More informations about this toot | More toots from zachleat@zachleat.com

Written by James Cridland on 2025-01-22 at 16:37

@snugug @zachleat Webp is supported on pretty well any device, excepting older iPhone. While it isn’t supported well in email clients, for the web it’s pretty well ubiquitous. My personal site just uses webp, with one solitary jpeg in the og:image field. (But I don’t support older iPhone…)

=> More informations about this toot | More toots from james@bne.social

Written by Jeremiah Lee on 2025-01-22 at 16:26

@zachleat AVIF is definitely worth it for me for the better compression and HDR support.

=> More informations about this toot | More toots from Jeremiah@alpaca.gold