I hopped from arch (2010-2019) to Nixos (2019-2023). I had my issues with it but being a functional programmer, I really liked the declarative style of configuring your OS. That was until last week. I decided to try out void Linux (musl). I’m happy with it so far.

Why did I switch?

  1. Nix is extremely slow and data intensive (compared to xbps). I mean sometimes 100-1000x or more. I know it is not a fair comparison because nix is doing much more. Even for small tweaks or dependency / toolchain update it’ll download/rebuild all packages. This would mean 3-10GB (or more) download on Nixos for something that is a few KB or MB on xbps.

  2. Everything is noticeably slower. My system used way more CPU and Ram even during idle. CPU was at 1-3% during idle and my battery life was 2 to 3.5h. Xfce idle ram usage was 1.5 GB on Nixos. On Void it’s around 0.5GB. I easily get 5-7h of battery life for my normal usage. It is 10h-12h if I am reading an ebook.

Nix disables a lot of compiler optimisations apparently for reproducibility. Maybe this is the reason?

  1. Just a lot of random bugs. Firefox would sometimes leak memory and hang. I have only 8 GB of ram. WiFi reconnecting all the time randomly. No such issues so far with void.

  2. Of course the abstractions and the language have a learning curve. It’s harder for a beginner to package or do something which is not already exposed as an option. (This wasn’t a big issue for me most of the time.)

For now, I’ll enjoy the speed and simplicity of void. It has less packages compared to nix but I have flatpak if needed. So far, I had to install only Android studio with it.

My verdict is to use Nixos for servers and shared dev environments. For desktop it’s probably not suitable for most.

  • featherfurl@lemmy.ml
    link
    fedilink
    arrow-up
    7
    ·
    edit-2
    1 year ago

    I also haven’t noticed a significant performance hit from using nixos on desktop coming from arch a few months ago. Nix definitely does a lot of stuff and that can chew through bandwidth at times, but overall I think the time saved from not compiling heaps of aur packages has outweighed the time lost to nix updating and maintaining the overall state of my system on every update.

    I tend to run relatively lightweight systems these days and haven’t really noticed sluggishness compared to an equivalent system on arch. My desktop environment has been sway on both for a while and this may account for my experience of a leaner and more reliable system on both, but it’s hard to say.

    I’d definitely want to investigate bandwidth optimization strats for nix if I was heavily constrained in that area, or possibly move to something where cpu and bandwidth constraints were given priority over reproducibility. For my current setup nixos has been a game changer on both desktop and server, but I only really have arch as a direct comparison.

    ( For context, my current desktop nixos systems are a 9 year old low-end cintiq, a 2017 dell optiplex 7050 minipc, and a steam deck. They all have ssds and at least 12gb of ram. All feel super snappy for everyday work with a web browser and a heap of open terminals and workspaces. )

    • 7aiOP
      link
      fedilink
      arrow-up
      1
      ·
      edit-2
      1 year ago

      Thank you that makes sense. When I get my hands on a more powerful machine and have less data constraints, I’ll try Nixos again. I do miss it sometimes 😆