AllHailTheSheep

my i7 3700k with 8gb of ddr3 handles multiple 4k streams just fine. it doesn’t transcode on the fly super well, I transcode everything to hevc before putting it on the server. it will happily transcode a single 4k stream in the rare case a client doesn’t want hevc.

point is, don’t target latest gen simply because it’s latest. 12th gen is definitely newer than I’d opt for, and while 16gb is cool, 32 is absolutely overkill for the use cases you listed (other than the llm, which others properly list as “don’t worry about it unless you have 3k in your back pocket to shell on gpus”)

whoops!! that’s an interesting overlap of issues! glad a new cable did the trick (and that jellyfin fixed that).

this is the way imo. I also took the time to disable internal authentication and hook up authentik via the ldap auth plugin. so for my users, I can text them a link, they set up their credentials and verify their email, then they can sign in to jellyfin with the same creds and it automatically creates a profile and everything.

best money I’ve ever spent

I’m so curious what the cable was and how it affected only ass/ssa subtitles?

take it out of that adapter. lots of people giving good tips but it’s not worth assuming anything about the drive till you plug it directly in. that’s a pretty high capacity drive, and shitty adapters can misbehave in fascinating ways.

faint clicking is normal from a drive of this size. bigger drives make bigger noises (especially those designed for use in enterprise environments).

when you plug it in, run smartctl -a on it and post it here. that’ll be definitive for what’s up before you try any of the more creative solutions here.

neat! very unique idea which is impressive these days

I have that ups! apc br1500g?

looks like an older nvidia workstation card. potentially a tesla k80 of the top of my head? the annoying thing with those cards is that they’re meant to be passively cooled in a temperature controlled environment, so looks like they added a fan to help keep it cool. those cards get mad hot otherwise.

not really. you can disable the default jellyfin login and force it to use ONLY ldap.

you totally can use ldap or oidc it just requires more setup. you just ensure jellyfin and your source of truth talk on their own subnet, docker can manage it all for you. ldap can be setup to be ldaps with ssl and never even leave the docker subnet anyways.

and yes I suppose you could rely on whitelists, but you’d have to manually add to the whitelist for every user, and god forbid if someone is traveling.

ldap auth will work on tvs

or use the ldap auth plugin with your source of truth, put it behind a reverse proxy, protect it with fail2ban and anubis. there are ways of exposing it safely.

thanks for posting this!

yeah agreed. I’m fully in support of an open source discord alternative, it’s sorely needed, but expecting to vibe code real time communication tells me exactly which programmers have worked with real time systems before. RTC is not something you vibe code successfully haha.

there’s a .cursor directory in there with a rule specifically telling it to not mark generated code as generated. I can’t see how that could ever cause any issues /s

really cool idea. ty for sharing!

dead ram definitely still happens, yes, but it’s exceedingly rare. I fix hundreds of PCs a year, and I maybe get one or two a year where the root cause is actually bad ram. more often it’s configuration issues or hardware implementation issues, for example the gigabyte x870 boards really don’t like XMP for some reason.

ecc doesn’t really have anything to do with whether a ram stick fails or not, it can help with misbehaving sticks but if a stick is dead it’s dead and ecc can’t help a dead region.

smartctl would be what your looking for even for ssds (although ssds fail quick enough that if smartctl catches something there’s a chance it’s already too late, smartd allows for scheduled tests and I’ve definitely saved data off of ssds because I had daily smart tests running that caught early failure).

I however strongly disagree with the hardware issue. there is no indication that this is hardware (honestly hardware accounts for VERY few issues like this, and RAM failing still happens but is 98% a thing of the past). diagnosing without any logs is a bit of a lost cause, we simply don’t have enough info, hopefully OP updates the post with the output of journalctl from the last boot.

musl is very cool and has very specific use cases. workstations are not one of them. you won’t be able to install drivers for gpus for example.