Newegg must really be desperate to get rid of those graphics cards now that GPU mining is useless.

  • @holdengreen
    link
    82 years ago

    Yes 28nm is a long node. It provides great value in cost and efficiency for companies who cannot afford to produce at a lower node or for products which don’t have the volume to justify buying a lower node.

    But maybe 14nm will be next since China is gaining the ability to mass produce it. Moore’s Law also seems to break down pretty bad after 14nm since transistors get more expensive and of course no one can afford to make 7nm chips aside from corporations like Apple, AMD, Samsung. In a long enough time frame this may adjust with EUV being normalized.

    It may happen sooner rather than later that China produces it’s first domestic EUV machines. Whatever ability China has to manufacture it’s own lithography machines will directly translate to the semiconductor security of Russia.

    Xi has announced publicly their interest in promoting semiconductors with alternative materials such as carbon. (I read the second volume of Xi’s book.) I believe that this will be a very big deal eventually.

      • @holdengreen
        link
        12 years ago

        Maybe if u can get a good deal on it. Running stuff like that off of a battery is a pain. It will be better when China goes full riscv.

        But I have extra ATX parts sitting around so maybe I will be tempted to build an exotic Chinese system at some time.

          • @holdengreen
            link
            22 years ago

            That doesn’t change that riscv is still generally better if you care about efficiency and cost.

            x86 needs to do all sorts of crap to stay compliant and compatible. It’s done at the die level.

              • @holdengreen
                link
                32 years ago

                I guess. I’m excited to start programming on the Lichee Allwinner D1 dev kit I just ordered for $27 after shipping and tax. (that my grandma offered to buy me)

                riscv has still got to be a pretty useful micro architecture compared to ARM which still ought to eat more transistors and be less pleasant to program for on the ultra embedded.

                alibaba is already using it for HPC so it’s got to be applicable