• ogeist@lemmy.world
    link
    fedilink
    English
    arrow-up
    21
    ·
    8 hours ago

    I’ve got the feeling that GPU development is plateauing, new flagships are consuming an immense amount of power and the sizes are humongous. I do give DLSS, Local-AI and similar technologies the benefit of doubt but is just not there yet. GPUs should be more efficient and improve in other ways.

    • PM_Your_Nudes_Please@lemmy.world
      link
      fedilink
      English
      arrow-up
      21
      ·
      7 hours ago

      I’ve said for a while that AMD will eventually eclipse all of the competition, simply because their design methodology is so different compared to the others. Intel has historically relied on simply cramming more into the same space. But they’re reaching theoretical limits on how small their designs can be; They’re being limited by things like atom size and the speed of light across the distance of the chip. But AMD has historically used the same dies for as long as possible, and relied on improving their efficiency to get gains instead. They were historically a generation (or even two) behind Intel in terms of pure hardware power, but still managed to compete because they used the chips more efficiently. As AMD also begins to approach those theoretical limits, I think they’ll do a much better job of actually eking out more computing power.

      And the same goes for GPUs. With Nvidia recently resorting to the “just make it bigger and give it more power” design philosophy, it likely means they’re also reaching theoretical limitations.

      • Redredme@lemmy.world
        link
        fedilink
        English
        arrow-up
        3
        arrow-down
        1
        ·
        3 hours ago

        AMD never used chips “more efficiently”. They hit gold with the RYZEN design but everything before since Athlon was horrible and more useful as a room heater. And before athlon it was even worse. The k6/k6-2 where funny little buggers extending the life of ziff7 but it lacked a lot of features and dont get me started about their dx4/5 stuff which frequently died in spectacular manners.

        Ryzen works because of chiplets and the stacking of the cache. Add some very clever stuff in the pipeline which I don’t presume to understand and the magic is complete. AMD is beating intel at it’s own game: it’s Ticks and tocks are way better and most important : executable. And that is something Intel hasn’t been able to really do for several years. It only now seems to be returning.

        And lets not forget the usb problems with ryzen 2/3 and the memory compatibility woes of ryzen’s past and some say: present. Ryzen is good but its not “clean”.

        In GPU design AMD clearly does the same but executes worse then nvidia. 9070 cant even match its own predecessor, 7900xtx is again a room heater and is anything but efficient. And lets not talk about what came before. 6xxx series where good enough but troublesome for some and radeon 7 was a complete a shitfest.

        Now, with 90 70 AMD once again, for the umpteenth time, promises that the generation after will fix all its woes. That that can compete with Nvidia.

        Trouble is, they’ve been saying that for over a decade.

        Intel is the one looking at GPU design differently. The only question is: will they continue or axe the division now gelsinger is gone. which would be monunentally stupid but if we can count on 1 thing then its the horrible shortsightness of corporate America. Especially when wall street is involved. And with intel, wall street is heavily involved. Vultures are circling.