• Paragone@lemmy.world
    link
    fedilink
    English
    arrow-up
    0
    ·
    27 days ago

    Years of afterburner-level development.

    chip-development is the worst quagmire of insane-numbers-of-details there is, in my view.

    https://store.steampowered.com/app/3033160/Hard_Chip_Demo/

    When you see just how many detals have to be got exactly right to make a 4 bit processor … & then you realize that GPU’s are 64-bit, & there are … what, 64 processors on the chip? something like that …

    & every single 1 needs its own memory area, & every memory area needs its own controller/manager…

    etc…

    the amount of details to accomplish ANYthing is insane, in chip-design, but to do real full-size useful IC’s?

    You need supercomputers dedicated to the calculating.

    There was a good video on yt about chip-engineering…

    “it’s basically magic” is the sentiment that the speaker tried to oppose, but had to admit was true.

    When you’re using optics to shine photons through a mask, so that the target-chip gets light where you want material to either stay or be removed…

    but you’re using masks & lenses to concentrate light down to features which are … the size of 20 hydrogen-atoms??

    Then quantum-mechanics is monkeying your optics.

    That video had a bit where the guy showed us a strange looking pattern, & said roughly “you make the mask look like that, & then you get a straight line on the chip: that shape was calculated by a supercomputer”.

    EVERY FEATURE-SHAPE has to be calculated by supercomputer, due to the wave/particle weirdnesses, & you’re cramming billions & billions of 'em together in there… ( yes, you can re-use shapes so long as the nearby “geography” is the same as the other-instances, but it’s still insane: you have to calculate each unique instance, & you have to engineer to reduce unique-relative-geometries, so as to make your shape-calculating more affordable? )


    It’d take an army-level budget, it’d take years, it’d take the best minds, it’d take oceans-of-compute-power, & it’d be worth every cent, if it kept the EU alive after the Regional Consolidation Time is finished, in 7-ish years or so.

    NOT doing it will be fatal for the EU.

    _ /\ _