Login
You're viewing the functional.cafe public feed.
  • Nov 21, 2026, 11:51 PM

    @marginalia

    This is exactly what our National labs used to do.
    One of them, Oak Ridge in Tennessee did something amazing as a side effect of some very basic research.

    Back in the late 90's and early 00's there was a race for cheaper supercomputers. Beowolf clusters were a thing, as were COWS (clusters of workstations) and LOBOS (lots of boxes on shelves).

    1/4

    💬 1🔄 0⭐ 2

Replies

  • Nov 21, 2026, 11:55 PM

    @marginalia
    The researchers at ORNL were trying to model the trajectories of particles in nuclear reactions. It could have some practical uses but was mainly in support of theoretical Physics research.

    Pretty out there stuff in terms of ever producing vast wealth from the investment.

    The researchers needed more computing power on a low budget and they figured out something

    2/2

    💬 1🔄 0⭐ 1
  • Nov 22, 2026, 12:00 AM

    @marginalia

    A team at ORNL figured out how to update the code in cheap commodity graphics cards so they would run the math calculations they needed instead of the graphical calculations the GPUs were designed to run.

    Instead of buying tons of servers, or even server blades that cost many times more than early GPUs, they could use server racks full of graphics cards to build a cheaper supercomputer.
    3/4

    💬 2🔄 0⭐ 1
  • Nov 22, 2026, 12:05 AM

    @marginalia

    And that's how Nvdia became a multi billion dollar company.

    No one was researching how to make a billion dollar idea. They were researching how particles obey the rules of quantum mechanics. They could have been biologists studying protein, or entomologists studying mosquitoes.

    Who knows?

    Nvdia was just a side effect of some esoteric research.

    Just like the way we learned so much about human biology during the Apollo program, as a side effect.

    Basic research is cool!
    4/4

    💬 0🔄 0⭐ 3
  • 💬 0🔄 0⭐ 1