• Hyacin (He/Him)@lemmy.ml
    link
    fedilink
    English
    arrow-up
    25
    ·
    edit-2
    1 day ago

    I mostly agree, and did the same with my second gen lab build - instead of shiny new NUCs like I had used round 1, I bought old off lease Dell Xeon boxes. SO MANY PROS -

    • Got them up to 14c/28t each
    • They can take GPUs and actually do heavy transcoding/ML work
    • They can take up to like, 128GB of memory, which is GREAT when they’re all hypervisors

    The downsides can’t be denied though -

    • Even without the GPUs and beefed up CPUs, they are power hogs - the CPU alone uses more than an ENTIRE NUC
    • They run HOT
    • They run LOUD

    The same holds true for off-lease SFF stuff, Lenovo and the likes …

    So while reuse/repurpose is absolutely of the utmost importance, no question - when it comes to technology and how quickly it advances and miniaturizes, a thorough and logical pros/cons list is often required.

    I’d add another option though - if you do need what a Pi brings to the table - do you really need a shiny new Pi 5? Is it possible a used Pi 3 or Pi 4 would do the trick, and check the reuse box?

    • AndrasKrigare@beehaw.org
      link
      fedilink
      arrow-up
      5
      ·
      1 day ago

      The power aspect is a lot bigger of a factor than I would have thought. I had an old computer I was going to use as a server for Foundry that I could keep up all the time, but when I measured its wattage and did the math, it would cost me $20 a month to keep on. A pi costs like $2 to keep running, so it paid for itself pretty quick

    • areyouevenreal@lemm.ee
      link
      fedilink
      arrow-up
      2
      ·
      22 hours ago

      RPis aren’t energy efficient either. Any situation where you are thinking of putting more than one of them in a cluster you should just buy mini PCs instead.