• ☆ Yσɠƚԋσʂ ☆@lemmygrad.mlOP
      link
      fedilink
      English
      arrow-up
      14
      ·
      6 days ago

      China has a pretty good track record taking new tech like this to market. Just look at all the battery tech advances we’ve seen over the past few years.

        • ☆ Yσɠƚԋσʂ ☆@lemmygrad.mlOP
          link
          fedilink
          English
          arrow-up
          10
          ·
          6 days ago

          Sure, I’m just pointing out that I’m less dismissive of new breakthroughs coming out of China. There’s huge government investment in supporting tech, and that addresses the problem of the cost of taking something that works in a lab and scaling it to commercial output.

    • polpotkin [none/use name]@hexbear.net
      link
      fedilink
      English
      arrow-up
      5
      ·
      5 days ago

      Technology changes a lot from R&D announcement to actual commercial production. Think of it as ‘we were able to prove this could be a viable direction’ rather than a finished product that just needs to be scaled up.

  • TankieTanuki [he/him]@hexbear.net
    link
    fedilink
    English
    arrow-up
    24
    ·
    edit-2
    6 days ago

    PoX could be the key to unlocking performance bottlenecks caused by storage limitations in AI hardware

    Is AI all this publication can think about?

    • ☂️-@lemmy.ml
      link
      fedilink
      English
      arrow-up
      6
      ·
      5 days ago

      yeah, thatd literally revolutionize computers, if true and applicable at scale.

  • peeonyou [he/him]@hexbear.net
    link
    fedilink
    English
    arrow-up
    8
    ·
    6 days ago

    I missed the “flash” part of the article and saw it was talking about writing 25 gigabits/s and was wondering wtf… we have faster ram than that? But flash? Whoa buddy!

    • ☆ Yσɠƚԋσʂ ☆@lemmygrad.mlOP
      link
      fedilink
      English
      arrow-up
      18
      ·
      6 days ago

      This could be absolutely revolutionary because it eliminates the need to have persistent storage and the bus. This way you can just have RAM that’s non-volatile. Even if it’s slower than fastest RAM the benefit of eliminating the need for the bus completely offsets that in my opinion. This would be a similar benefit as what we see with SoC architecture like M1 where memory can be shared between the CPU and GPU without having to copy things over the bus between regular memory and vram.