• makeasnek@lemmy.ml
    link
    fedilink
    arrow-up
    21
    arrow-down
    3
    ·
    edit-2
    1 year ago

    These things are great for [email protected] often time leagues more efficient per watt in terms of computation than regular PCs. I have a couple of 'em working on cancer research and computing to develop an open-source patent-free covid antiviral. You don’t need a PhD to make a difference, all you need is a processor :)

    • bionicjoey@lemmy.ca
      link
      fedilink
      arrow-up
      20
      arrow-down
      3
      ·
      1 year ago

      If they were more efficient per watt for scientific computing, you’d hear about researchers building HPC clusters from them.

      • makeasnek@lemmy.ml
        link
        fedilink
        arrow-up
        4
        ·
        edit-2
        1 year ago

        If they were more efficient per watt for scientific computing, you’d hear about researchers building HPC clusters from them.

        Efficiency per watt is not the same as total cost of ownership. Pis are expensive for the amount of compute you get from them in total, but the compute itself is efficient per watt. You would need at least a dozen Pis to rival the latest CPU processors in terms of total output, a dozen Pis is more expensive to buy than a single CPU.

    • towerful@programming.dev
      link
      fedilink
      arrow-up
      9
      ·
      1 year ago

      I did a quick Google.
      https://web.eece.maine.edu/~vweaver/group/green_machines.html

      Is the best actual test data I can find. It uses a physical power meter, so it’s full system (not TDP or self reporting power consumption).
      And it’s a few years out of date.
      Seems like Apple silicon is the winner (and will probably continue to be).
      The Xeon that beats the rpi4 for GFLOPS/watt is an e5v3, which was launched in 2013 and EOL in 2021.
      So there will absolutely be some new Xeon CPUs that will perform better.

      However, for a $50 device, it’s probably the best GFLOPS/watt/$ from what little empirical data I can find