Members who reacted to this

All (3) Like Like (3)

    1. Like

      Touko White

      (not)Banned
      • Messages
        687
      • Blogs
        1
      • Blog entries
        4
      • Blog entry comments
        319
      • Reaction score
        1,487
      • Trophies
        0
    2. Like

      Aurora Wright

      Well-Known Member
      • Messages
        1,550
      • Blogs
        2
      • Blog entries
        5
      • Blog entry comments
        108
      • Reaction score
        4,163
      • Trophies
        3
  • Loading…
General chit-chat
Help Users
  • The Real Jdbye @ The Real Jdbye:
    the vram is one advantage when it comes to AI but ends up being slower even with that and really AI is the only use case that needs more than 12gb vram right now
  • Psionic Roshambo @ Psionic Roshambo:
    Interesting lol
  • Psionic Roshambo @ Psionic Roshambo:
    I think I watched a video where two games at 4K where eating just over 16GB of RAM and it's the one case where the 7900XT and XTX pulled ahead (minus RTX of course)
  • Psionic Roshambo @ Psionic Roshambo:
    So my opinion is that they could age a bit better in the future, and maybe AMD will continue improving them via drivers like they tend to do. No guarantee there but they have done it in the past. Just a feeling I have.
  • The Real Jdbye @ The Real Jdbye:
    cyberpunk at 4k without DLSS/fidelityfx *might* exceed 12gb
    +1
  • The Real Jdbye @ The Real Jdbye:
    but that game barely runs at native 4k
  • Psionic Roshambo @ Psionic Roshambo:
    I think it was some newer games and probably poorly optimized PS4 or PS5 ports
  • The Real Jdbye @ The Real Jdbye:
    they definitely will age better but i feel dlss might outweigh that since it looks about as good as native resolution and much less demanding
    +1
  • Psionic Roshambo @ Psionic Roshambo:
    When I played Cyberpunk on my old 2080 Ti it sucked lol
  • The Real Jdbye @ The Real Jdbye:
    AMD could introduce something comparable to DLSS but nvidia's got a lot more experience with that
  • The Real Jdbye @ The Real Jdbye:
    least amd 7xxx has tensor cores which the previous generations didn't so there is the potential for AI upscaling
  • Psionic Roshambo @ Psionic Roshambo:
    They have FSR or whatever it's called and yeah it's still not great
  • The Real Jdbye @ The Real Jdbye:
    so AMD seem to finally be starting to take AI seriously
  • Psionic Roshambo @ Psionic Roshambo:
    Oh yeah those new 8000 CPUs have AI cores built in that's interesting
  • Psionic Roshambo @ Psionic Roshambo:
    Maybe they plan on offloading to the CPU?
  • Psionic Roshambo @ Psionic Roshambo:
    Would be kinda cool to have the CPU and GPU working in random more
  • Psionic Roshambo @ Psionic Roshambo:
    Tandem even
  • The Real Jdbye @ The Real Jdbye:
    i think i heard of that, it's a good idea, shouldn't need a dedicated GPU just to run a LLM or video upscaling
  • The Real Jdbye @ The Real Jdbye:
    even the nvidia shield tv has AI video upscaling
  • The Real Jdbye @ The Real Jdbye:
    LLMs can be run on cpu anyway but it's quite slow
  • BakerMan @ BakerMan:
    Have you ever been beaten by a wet spaghetti noodle by your girlfriend because she has a twin sister, and you got confused and fucked her dad?
  • Psionic Roshambo @ Psionic Roshambo:
    I had a girlfriend who had a twin sister and they would mess with me constantly.... Until one chipped a tooth then finally I could tell them apart.... Lol
  • Psionic Roshambo @ Psionic Roshambo:
    They would have the same hair style the same clothes everything... Really messed with my head lol
  • Psionic Roshambo @ Psionic Roshambo:
    @The Real Jdbye, I could see AMD trying to pull off the CPU GPU tandem thing, would be a way to maybe close the gap a bit with Nvidia. Plus it would kinda put Nvidia at a future disadvantage since Nvidia can't make X86/64 CPUs? Intel and AMD licensing issues... I wonder how much that has held back innovation.
    Psionic Roshambo @ Psionic Roshambo: @The Real Jdbye, I could see AMD trying to pull off the CPU GPU tandem thing, would be a way to...