←back to thread

294 points AgaoAnar | 1 comments | | HN request time: 0s | source
Show context
2OEH8eoCRo0 ◴[] No.45108785[source]
Have you measured the power consumption of your "high power consumption" server?
replies(3): >>45109039 #>>45109962 #>>45110348 #
jeffbee ◴[] No.45109039[source]
Seriously. You can get running states around 1W these days.
replies(3): >>45109835 #>>45110165 #>>45110346 #
macawfish ◴[] No.45109835[source]
What to do about this? https://forums.developer.nvidia.com/t/phantom-rtx-5090-high-...
replies(3): >>45109936 #>>45110761 #>>45111296 #
jeffbee ◴[] No.45109936[source]
I suggest not putting a ridiculous GPU in your backup server.
replies(1): >>45111446 #
overfeed ◴[] No.45111446{3}[source]
Not everyone's "home server" is a backup server - some are for AI experiments.
replies(1): >>45113358 #
silon42 ◴[] No.45113358{4}[source]
That one should be Wake-On-LAN
replies(1): >>45117521 #
1. overfeed ◴[] No.45117521{5}[source]
I wouldn't want to fiddle with WoL just to get my self-hosted LLM-assisted code-completion working when I pickup my old and underpowered laptop on the couch for some quick hacking. Wake-on-any-network-activity would be perfect, and has superior UX.