• pedz@lemmy.ca
    link
    fedilink
    arrow-up
    84
    ·
    2 days ago

    I’m always a bit amazed of how things have progressed and on what Linux can still run.

    This is an extreme example, but it’s also possible to run a modern Linux OS on SBCs like a Raspberry Pi Zero, and still have something somewhat usable depending on your needs.

    To have a computer half the size of a credit card with more RAM than my full tower rig from 2001 is amazing. And it can even run software from that era with dosbox or wine.

    My 15 years old laptop is still supported and can still read 1080p on YouTube, using Linux.

    Linux devs just recently decided to drop support for 486 CPUs and some early Pentiums.

    There’s just no competition.

    • henfredemars@infosec.pub
      link
      fedilink
      English
      arrow-up
      21
      ·
      2 days ago

      Pretty much the only place it doesn’t run is where you have hard real-time requirements and on extremely small embedded micro controllers.

        • henfredemars@infosec.pub
          link
          fedilink
          English
          arrow-up
          14
          ·
          2 days ago

          Sort of. There’s realtime builds, but the Linux definition of real time is more relaxed than dedicated RTOS’s in exchange for a much more feature-filled OS. You should not use Linux if people could die when you miss a deadline. You want a simple system where it’s easier to prove that can never happen.

        • Colloidal@programming.dev
          link
          fedilink
          English
          arrow-up
          17
          ·
          edit-2
          2 days ago

          Technically yes. But it can’t support many hard real-time use cases. For that you need a true RTOS, thought from the ground up for that purpose. Something like VxWorks, QNX, some flavors of L4.

    • Cenzorrll@lemmy.world
      link
      fedilink
      arrow-up
      9
      ·
      2 days ago

      My laptop with 4gb of ram and atom processor came with Windows 10 (I purchased it intending to put Linux on it). I tried it out just to see how W10 worked on it and it was absolutely excruciating and borderline unusable, and I was coming from a Pentium M with 512mb of ram. I have no idea how anyone would have thought it could be a functional system running that.

      • biggerbogboy@sh.itjust.works
        link
        fedilink
        arrow-up
        3
        ·
        2 days ago

        I cannot imagine how disgustingly slow your college ones are compared to my high school’s ones, which are 16gb ram, i7 boxes running windows 11. They are pretty damn slow even after they recently upgraded them a month ago.

        For most things, I just use my laptop since it stutters much less, but using pycharm on my laptop is pretty much impossible due to its lack of specs, so I begrudgingly use the PC.

  • Ephera@lemmy.ml
    cake
    link
    fedilink
    English
    arrow-up
    52
    ·
    2 days ago

    At $DAYJOB, we’ve been working on a service which uses Raspberry Pis as edge devices. And our product manager – bless him – has made sure we’d have enough hardware budget and wanted to buy only Raspberry Pi 5, so we’d have really good performance.

    And I think, we really befuddled him with our reaction, because you know, normally devs won’t say no to good hardware, but because our software happens to be efficient and Linux is efficient, we’ve just been like, eh, a Pi 3B+ is already a lot beefier than we need it.
    We had to explain that to him like five times before he actually started to believe it. 🙃

      • Ephera@lemmy.ml
        cake
        link
        fedilink
        English
        arrow-up
        18
        ·
        2 days ago

        Yeah, I considered explaining that differently, but figured it doesn’t really matter for the story. 😅
        It’s “edge” basically in the sense that it’s on-the-edge towards the physical world: https://en.wikipedia.org/wiki/Edge_computing

        In our case, there’s some dumb devices, which wouldn’t be able to talk across the internet on their own, so we put Raspberry Pis next to them to hook them up to the internet. In other words, the Raspberry Pis just push network packages through, they’re not going to be crunching numbers or whatever.

    • DeltaWingDragon@sh.itjust.works
      link
      fedilink
      arrow-up
      6
      ·
      1 day ago

      Gentoo users be like: Your system is underperformant and unoptimized. Have you tried Gentoo, and compiling all the packages?

      Other users be like:
      Have you tried Slackware?
      Have you tried Damn Small Linux?
      Have you tried Linux From Scratch?
      Have you tried hand-optimizing a 386BSD system with a Mach kernel? OOPS, that’s too far!

    • silly goose meekah@lemmy.world
      link
      fedilink
      arrow-up
      10
      ·
      edit-2
      2 days ago

      Honestly making sure my PC only has software installed that I actually use is important to me because my internet is super slow. It took me over 1h yesterday to do a system update. I don’t even wanna know how bad it would be if I had more than 2k packages, it’s already bad enough now with 1.1k

    • bruhduh@lemmy.world
      link
      fedilink
      arrow-up
      2
      ·
      2 days ago

      I once installed whole blackarch repo in my machine, so I’m not sure if my 90gb installation was bloated or not

    • DickFiasco@lemm.ee
      link
      fedilink
      arrow-up
      66
      arrow-down
      2
      ·
      2 days ago

      Right? 2 GB for 64-bit Windows 10? The start button alone probably needs more RAM than that.

      • MonkeMischief@lemmy.today
        link
        fedilink
        arrow-up
        2
        ·
        23 hours ago

        The start button alone probably needs more RAM than that.

        Nah, just all the ad telemetry gathered whenever the user hits the start button!

      • Sestren@lemmy.world
        link
        fedilink
        arrow-up
        21
        ·
        2 days ago

        What else are you going to use your RAM for if not generating advertisement revenue for shareholders.

  • ms.lane@lemmy.world
    link
    fedilink
    English
    arrow-up
    13
    ·
    2 days ago

    Linux on system RAM: I sip

    Linux on VRAM post-Wayland: BIG GULPS

    I used to be able to run a desktop on less than 1GB VRAM, now with 16GB it fills up starts misbehaving.

    • ryannathans@aussie.zone
      link
      fedilink
      arrow-up
      18
      ·
      2 days ago

      Ya might wanna check that because it’s not normal

      I play star citizen in Wayland and don’t even get close to 16GB vram utilisation

    • tal@lemmy.today
      link
      fedilink
      English
      arrow-up
      6
      ·
      edit-2
      2 days ago

      I have a 24GB VRAM video card, and the only time I’ve noticed it using over 16GB is when I’m doing LLM stuff. I don’t think that just running under Wayland has ever used much — it’s games that ask for video memory.

      Though for LLM stuff, I agree heartily. If I could get a video card with 128GB of VRAM (well, for a not-completely-insane price) I would. You can always use more VRAM there.

      But that’s not really due to Wayland.

      • ms.lane@lemmy.world
        link
        fedilink
        English
        arrow-up
        4
        ·
        2 days ago

        The problem is actually kwin-wayland, kwin5 on X11 was very frgual, where both kwin5 and kwin6 on Wayland eat vram for breakfast. Right now my desktop is using 4535MiB, the biggest user being plasmashell itself at 1120MiB

        It’s not actually a wayland problem though and If I really wanted I’d of dropped back to X11 when I was still running plasma5 - but Wayland brings more solutions to the table than problems.

        • tal@lemmy.today
          link
          fedilink
          English
          arrow-up
          3
          ·
          edit-2
          2 days ago

          https://lemmy.today/post/30379016/16636854

          According to a quick search I did, there is (was?) an issue causing excessive memory usage by Wayland compositors with Nvidia’s driver; there’s a manual fix listed at my link. Is there any chance that you’re using Nvidia hardware?

      • Valmond@lemmy.world
        link
        fedilink
        arrow-up
        2
        ·
        2 days ago

        I tried to try llms but my 4GB VRAM is just not at all enough, what’s the minimum you need in your opinion?

        • tal@lemmy.today
          link
          fedilink
          English
          arrow-up
          4
          ·
          edit-2
          2 days ago

          I tried to try llms but my 4GB VRAM is just not at all enough, what’s the minimum you need in your opinion?

          For just using Wayland, 4GB should be fine. I’m using, uh, 801M right now, and it doesn’t change that much. I think I’ve seen it get over 1GB, not to 2GB.

          For current 3D games, yeah, I agree, 4GB is very much not enough. If I were getting a card to play 3D games on in 2025, I’d probably aim for 16 GB. Could live with 12GB. 8GB is IMHO not what I would go with for PC gaming in 2025. Lots of stuff can run, but games might need to have texture resolution turned down or something.

          For LLM stuff, the sky is the limit. I mean, right now, I have image generation models on my computer that will blow past that, need to live partially in main memory. And I can certainly do things, like generation of images at the full resolution of my screen without upscaling using more-memory-intensive models that will exhaust 24GB of VRAM.

          If you’re running into problems with 4GB just running normal Wayland apps, something is wrong.

          kagis

          I see several people using Nvidia GPUs claiming that various scenarios cause their VRAM to be consumed under Wayland. I’m using an AMD GPU on sway myself, so maybe that could be a factor.

          https://forums.developer.nvidia.com/t/multiple-wayland-compositors-not-freeing-vram-after-resizing-windows/307939/12

          https://github.com/NVIDIA/egl-wayland/issues/126#issuecomment-2379945259

          This is talking about reports of two separate problems that Nvidia is working on related to VRAM usage under wayland. The guy here, one of the NVidia Linux driver guys, is working on one, says that there’s a manual fix for one of them, Wayland compositors using a lot of VRAM, and gives it in his comment.

          He also says that there are reports of some other issue with Xwayland apps (X11 apps running under Wayland) using a lot of memory, and he hasn’t been able to identify that problem.

          He made that post nine months ago, so I’d think that they may have rolled the fix he was talking about out to users. I do see several users saying that the manual fix solved their issue at the above links.