• 6 Posts
  • 1.06K Comments
Joined 2 years ago
cake
Cake day: June 13th, 2023

help-circle








  • “Nuclear is expensive” is a post-hoc circular argument, though. Nuclear has become expensive now because the hysteria about it, and accompanying massive political opposition, lawsuits, and paranoid regulations mandating massive overkill factors of safety, have succeeded in their goal of driving up the cost to the point of non-viability.

    But back in the '60s, before the backlash had a chance to get going, we were building tons of nuclear plants at such cost-effectiveness that the economic worry was that the power they would provide would be “too cheap to meter.” And guess what: with very few exceptions, those power plants were, in fact, good enough in terms of safety and released no notable pollution throughout their entire operating lifetimes (or lifetimes so far, for the ones that are still in operation today).

    At this point, sure, you’re right: it doesn’t make sense to build nuclear in 2025 because it’s incompatible with our paranoid regulatory environment, and also because our capability has atrophied after a generation of not training nuclear engineers because there was no job market for it. But that’s an “us” problem, not a problem inherent to the technology.

    The world would’ve been massively better off if Greenpeace had stuck to saving the whales and we had built a trillion killowatt-hours worth of nuclear over the past 40 years instead of the gas-fired generators we actually built. Even at the cost of an additional Chernobyl or two.


  • I read OP’s question as him streaming from a Jellyfin server to this box, not using this box as a Jellyfin server itself. Could be wrong, though.

    Also, it’s my understanding that transcoding is 100% about hardware support for the codecs and that integrated graphics that have it (TL;DR: 12th gen Intel) are going to perform pretty much just as well as even a high-end discrete gaming GPU for that task.

    (I say “gaming” GPU because I was reading about the Arc Pro B50 the other day and it has two separate sets of transcoding hardware, so it presumably would actually perform better in terms of the number of simultaneous streams it could handle. But short of something like that, it apparently doesn’t make much difference.)




  • I hate it when Free Software installers present the GPL as if it’s an EULA. It’s not! You don’t have to agree to it to install the software!

    You only have to agree to it if you decide to do something that copyright law otherwise does not allow (e.g. redistribution of a modified version), and it is the act of doing that thing itself that signals your acceptance, no button-clicking necessary.






  • Nothin’ I’m running, that’s for sure!

    It’s not really that there are services that require that much processing power for a single request; it’s that it’s designed to handle normal requests for hundreds or thousands of users at once.

    I suppose that supporting 0.5TB of RAM means it could deal with quite a big LLM, but any sort of halfway-modern GPU would absolutely run circles around it in terms of tokens per second, on any model that fit in their VRAM.