• MudMan@kbin.social
    link
    fedilink
    arrow-up
    13
    arrow-down
    8
    ·
    11 months ago

    None of that makes any sense. “Western chips” all come from Taiwan in the first place. “Western designed chips” are also in laptops and mobile phones, including tons of Chinese devices, and that’s assuming you mean to include South Korea as “Western”, which is a bit of a stretch. Those are fundamentally interchangeable with military hardware. Nobody is putting 4090s and A100s in ICBMs.

    Make it make sense. What specific hardware is this stopping from getting to China and for what application?

    • bassomitron@lemmy.world
      link
      fedilink
      English
      arrow-up
      15
      arrow-down
      1
      ·
      edit-2
      11 months ago

      It doesn’t matter, anyway. Sanctions like this can be easily sidestepped by China going through series of proxy vendors across the world and still receive whatever it is that they want.

      Regardless, it’s not so much the 4090s the US government cares a lot about, but rather the giant data center TPUs. They included the 4090s etc in the ban because Chinese government can easily afford to buy thousands of them to network together to accomplish the same thing as DC TPUs.

      As for the military application of the chips: You could absolutely reengineer and use a 4090 GPU chipset in an ICBM, but no I don’t think that’s what they’re really concerned with them doing. I’m betting they’re more concerned with their cyber warfare and other espionage/surveillance capabilities, which a 4090 can easily be repurposed for (especially a lot of them).

      Edit: And I don’t disagree with your earlier point! I would guarantee there’s a good deal of corporate protectionism going on. I honestly wouldn’t even be surprised if that’s 100% the real reason behind it. I was just trying to provide a plausible explanation for why it could be considered a military justification.

      • MudMan@kbin.social
        link
        fedilink
        arrow-up
        4
        arrow-down
        7
        ·
        11 months ago

        Right, but that’s my point, compute is compute is compute. There are tensor acceleration cores in commercially available hardware dating back five years. They capped things above a specific performance threshold, is my understanding, but that just means you need more of the less powerful hardware, so all you’ve done is make things more expensive/less energy-efficient, but not block any specific application. Not in cheap, portable chips, not in huge industrial data center processors.

        So not particularly useful to stop cyberwarfare, not particularly useful to stop military applications. The only use I see is making commercial applications less competitive. Specifically on the training side of things.

        • vaultdweller013@sh.itjust.works
          link
          fedilink
          English
          arrow-up
          2
          ·
          11 months ago

          The goal is kinda to just fuck China overall. Also youre wrong about military application, just cause you can duck tape 500 PS2s together to achieve the same thing as my PC doesnt mean it will be as fast, efficient, or as small as my pc.

          Just using an example say an American General Dynamics AA missile weigh 400 pounds with the newest and best hardware, to achieve the same thing cludging together a bunch of older hardware for an equivalent Chinese missile may very well increase the weight to 450 pounds, which in turn can effect speed, maneuverability, and even explosive yield.

          Remember theres a reason nobody cludges together a bunch of vista era computers to try to match a modern PC on a practical level.