Deleted
Edit history
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running,
Synapse
PostgreSQL
Cinny
Arcane
qBittorrent
RDT-Client
Radarr
Sonarr
Prowlarr
Docker
Dropbear
Fail2ban
Tailscale
Syncthing
Argon One
TigerVNC
it sits under a GB
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 has AXV-512 (not that it’s particularly required for transcoding but it helps) so double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity mini or even GPT-OSS 20B could work) but the more parameters the slower it’ll run. OTOH, just more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running,
Synapse
PostgreSQL
Cinny
Arcane
qBittorrent
RDT-Client
Radarr
Sonarr
Prowlarr
Docker
Dropbear
Fail2ban
Tailscale
Syncthing
Argon One
TigerVNC
it sits under a GB
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 actually has AXV-512 not that it’s particularly required for transcoding but double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity mini or even GPT-OSS 20B could work) but the more parameters the slower it’ll run. OTOH, more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running,
Synapse
PostgreSQL
Cinny
Arcane
qBittorrent
RDT-Client
Radarr
Sonarr
Prowlarr
Docker
Dropbear
Fail2ban
Tailscale
Syncthing
Argon One
TigerVNC
it sits under a GB
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 actually has AXV-512 not that it’s particularly required for transcoding but double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity or event GPT-OSS 20B could work) but the more parameters for slower it’ll run. OTOH, more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running,
,
Synapse
PostgreSQL
Cinny
Arcane
qBittorrent
RDT-Client
Radarr
Sonarr
Prowlarr
Docker
Dropbear
Fail2ban
Tailscale
Syncthing
Argon One
TigerVNC
it sits under a GB
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 actually has AXV-512 not that it’s particularly required for transcoding but double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity or event GPT-OSS 20B could work) but the more parameters for slower it’ll run. OTOH, more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running,
, it sits under a GB
Synapse
PostgreSQL
Cinny
Arcane
qBittorrent
RDT-Client
Radarr
Sonarr
Prowlarr
Docker
Dropbear
Fail2ban
Tailscale
Syncthing
Argon One
TigerVNC
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 actually has AXV-512 not that it’s particularly required for transcoding but double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity or event GPT-OSS 20B could work) but the more parameters for slower it’ll run. OTOH, more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running, it sits at around 850mb
Synapse
PostgreSQL
Cinny
Arcane
qBittorrent
RDT-Client
Radarr
Sonarr
Prowlarr
Docker
Dropbear
Fail2ban
Tailscale
Syncthing
Argon One
TigerVNC
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 actually has AXV-512 not that it’s particularly required for transcoding but double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity or event GPT-OSS 20B could work) but the more parameters for slower it’ll run. OTOH, more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.
I can speak to this because one of my old servers was pretty much the machine you’re proposing.
Briefly, in terms of distros, choose what you like (within reason). I’ve always trended towards dietpi, which is extremely light. I actually just stood it up on the Raspberry Pi 4B last night. Headless and by itself it takes a whole whopping 300MB of RAM at idle. Even with all this running, it sits at around 850mb
- Synapse PostgreSQL Cinny Arcane qBittorrent RDT-Client Radarr Sonarr Prowlarr Docker Dropbear Fail2ban Tailscale Syncthing Argon One TigerVNC
Do your own research, etc.
Pushing 4k streams, it’s not overkill at all, but by the sounds of it, you’re meaning to transcode on the fly? Be aware of AXV-512 support issues etc in Linux vs diff chips. Actually better make sure that your i5 actually has AXV-512 not that it’s particularly required for transcoding but double check.
Yes, you can run reasonable LLMs on that machine, though without a GPU, you’ll need to investigate exactly which backend runner best supports your CPU.
If you’re looking at a dense model on that rig, consider something in the 4 to 9 billion parameter range. Most comfortably, something like a Qwen3.5-4B, Gemma-4-E4B, MiniCPM or Tiny Ling…or a larger MoE (I like grug 35b-a3b, but Arcee Trinity or event GPT-OSS 20B could work) but the more parameters for slower it’ll run. OTOH, more params =! always better. See -
Gaming wise, obviously you want something that can run Proton, wine, DOSBox, etc. I can’t claim that my pet distro runs all of those OOB because I’ve never tried, but it’s been my general experience that Ubuntu variants seem to do gaming best.