Conversation

Notices

1. Embed this notice
   Amanda (amanda@social.spejset.org)'s status on Wednesday, 01-Nov-2023 22:49:10 JST Amanda

   Does anyone have numbers on computer power efficiency v. manufacturing energy? Ie when, if at any point, is a computer so old it’s actually a net negative to keep using it rather than upgrading?

   #RedueReuseRecycle

   • Embed this notice
     pettter (pettter@mastodon.acc.umu.se)'s status on Wednesday, 01-Nov-2023 22:48:50 JST pettter

     in reply to

     • maswan

     @amanda @maswan Yeah, SE2 has, like, the second lowest CO2e/kWh in the world or something like that (SE1 being the top).

   • Embed this notice
     Amanda (amanda@social.spejset.org)'s status on Wednesday, 01-Nov-2023 22:48:54 JST Amanda

     in reply to

     • maswan

     @maswan thanks! This actually changes my calculus somewhat

   • Embed this notice
     maswan (maswan@mastodon.acc.sunet.se)'s status on Wednesday, 01-Nov-2023 22:48:57 JST maswan

     in reply to

     @amanda
     Probably, yes, and if looking just at CO2 emissions. It gets different if your electricity has a large share of fossil emissions in it though.

   • Embed this notice
     Amanda (amanda@social.spejset.org)'s status on Wednesday, 01-Nov-2023 22:49:02 JST Amanda

     in reply to

     • maswan

     @maswan that sounds like what I’d have expected for a lower estimate. I wonder how this compares for home computing which isn’t dominated by compute performance. It sounds like “If the computer boots at all it’s worth using it over replacing it” if I extrapolate a bit

   • Embed this notice
     maswan (maswan@mastodon.acc.sunet.se)'s status on Wednesday, 01-Nov-2023 22:49:04 JST maswan

     in reply to

     @amanda Having recently done the numbers for servers for scientific computing, the manufacturing seems to be 1-2 tons of CO2e. For running computers in SE2, assuming the local power mix emissions (11-15 g/kWh), the manufacturing emissions totally dominates for both 5 and 10 year lifespans even when running them flat out 24/7.

     Money is a different matter, where 5-7 years lifetime is more reasonable since you get 1-2 orders of magnitude more compute per kW.

     Slides: https://indico.cern.ch/event/1289243/contributions/5583080/attachments/2738397/4762827/20231020-CarbonNegativeComputing.pdf

   • Embed this notice
     pettter (pettter@mastodon.acc.umu.se)'s status on Wednesday, 01-Nov-2023 23:41:49 JST pettter

     in reply to

     • maswan

     @maswan Gotcha, so basically places like Poland, China, South Africa and Australia? @amanda

   • Embed this notice
     maswan (maswan@mastodon.acc.sunet.se)'s status on Wednesday, 01-Nov-2023 23:41:50 JST maswan

     in reply to

     • pettter

     @amanda
     SE3 is only 2-4 times as bad on average, which is still 10% of fossil gas (400-500 g/kWh is where you get same emissions from manufacture and use somewhere between 3-10 years).
     @pettter

   • Embed this notice
     Amanda (amanda@social.spejset.org)'s status on Wednesday, 01-Nov-2023 23:41:51 JST Amanda

     in reply to

     • pettter
     • maswan

     @pettter @maswan I think I’m in SE3 but let’s assume I’m doing all I can etc

   • Embed this notice
     pettter (pettter@mastodon.acc.umu.se)'s status on Wednesday, 01-Nov-2023 23:43:40 JST pettter

     in reply to

     • maswan

     @maswan Or, looking at the current map, most of the USA. @amanda