[–] 24 points 1 week ago (2 children)
  1. heat (dissipation)
  2. power
  3. network bandwidth
  4. cosmic rays (shielding)
  5. part failures (with so many parts, there’s always something broken
  6. eol parts (lifecycling in 3-5 yrs avg)

They are all challenges that can be overcome. But at what cost. Consensus is that economically it seems hilariously unviable. Datacenters tend to have to make their money back within about 3 years. The 3 year make money back issue is also why I believe the AI DC buildout is such a high risk gamble/bubble. In a year or two, all that hardware they made a year ago will need to be replaced. You just redo the whole investment again…

  • source
  • [–] 147 points 2 months ago (4 children)

    Seems like the real story here is one guy maintaining a critical infrastructure begging for volunteers to help out and nobody steps up. So yeah, I really don’t like what he does. And also the OSS community is failing him.

  • source
  • [–] 44 points 4 months ago (4 children)

    Between Framework at least trying to build an open ecosystem, there are very very few other options out there. Yes they can probably do better, still they are effectively the only decent open option (for me at least) at a reasonable price and quality.

  • source
  • parent
  • context
  • view more: next ›