• real_squids@sopuli.xyz
    link
    fedilink
    English
    arrow-up
    35
    ·
    2 days ago

    You can pry my “double” slot GPU out of my cold dead hands (good luck trying to run away with it, it’s heavy as fuck and needs a supporting post)

    • autriyo@feddit.org
      link
      fedilink
      English
      arrow-up
      17
      ·
      2 days ago

      Tbh, the whole card format feels very legacy, even for my Vega 56 “dual slot” card, and that thing “only” consumes ~230W.

      If ppl back then could’ve foreseen what obscenely power hungry parts would be shoehorned into the expansion card format, they probably would’ve chosen a different approach for GPUs specifically.

      • kibiz0r@midwest.social
        link
        fedilink
        English
        arrow-up
        11
        ·
        2 days ago

        iGPUs should have been a better option, but they were hamstrung by PCI conventions and graphics APIs favoring discrete VRAM.

        (Just look at how x86 SoC consoles run circles around similar-spec PCs.)

        I’m hoping that ARM is a chance to reset.

        • autriyo@feddit.org
          link
          fedilink
          English
          arrow-up
          3
          ·
          2 days ago

          I do like the modularity of discrete GPUs though.

          But a cooling setup similar to CPUs would’ve been better for airflow.

          • kibiz0r@midwest.social
            link
            fedilink
            English
            arrow-up
            5
            ·
            2 days ago

            Modularity is nice — both for personal preference reasons and incentivizing-market-competition reasons — but it does come at a cost.

            The thing is: even in our modular world right now, you don’t really have many choices. Two CPU companies, three GPU companies (two of them being the same as the CPU companies)…

            We could someday have a world where PC hardware is technically less modular than it is today but consumers have more choices in the marketplace than they do today.

          • real_squids@sopuli.xyz
            link
            fedilink
            English
            arrow-up
            1
            ·
            edit-2
            2 days ago

            Kinda hard to do when so many GPU vendors slap their memory and power circuits all over the place. Even if the die is in the same place cooler manufacturers would need to test fit a bajillion models, and on top of that they’d need insane R&D budgets to keep up with new additions, sometimes coming years after the original gpu comes out

            • autriyo@feddit.org
              link
              fedilink
              English
              arrow-up
              2
              ·
              1 day ago

              My concerns were more about the airflow path and less about actually interchangeable coolers, although those are a thing.

              Like it just feels wrong to blast the air into a solid PCB… That’s kind of solved with flow through designs, but a tower style cooler would probably be less noisy.

      • real_squids@sopuli.xyz
        link
        fedilink
        English
        arrow-up
        2
        ·
        edit-2
        2 days ago

        I was joking about it being 2.5 slots, tbf most modern cards should be triple slots. Mine is 300W and it’s pretty chunky to stay below 60C, best option for big cards is a horizontal mobo imo

        edit: unpopular opinion but I’d rather have a chunky card that stays cool as fuck than a slim one, that’s why I picked up the Nitro when I had a 6650XT