If you’ve been a part of the PC geek race for as long as I have, chances are you’ve upgraded your gaming machine enough times to have some spare hardware lying around. Or maybe you skipped the Ship of Theseus situation altogether by replacing your entire system with a shiny new PC. Either way, if you’re not actively using your old gaming buddy, there are plenty of ways to breathe new life into it – like turning it into a home server.
Interestingly, if you check the minimum hardware requirements for Proxmox, TrueNAS and other server platforms, you will realize that they can run on any old computer, including dinosaur systems. But let me tell you this: overclocking your PC for server tasks is actually a good thing, as it can allow you to work on projects that a budget system or cheap mini-PC would struggle to handle.
Gaming machines have enough system resources to power dozens of virtual guests
They may be outdated for modern names, but LXC and VMs are fair game
Unlike triple-A titles and typical eSports games where you need faster clock speeds to increase FPS and stay ahead of the curve, CPU performance matters little in home labs. Actually, the number of cores and threads is more important than boost clocks, but as long as you have a 6-core (or heck, even 4-core) processor, your machine can easily handle multiple containers and multiple VMs (especially if you go the CLI route). The source? Regards, who built a PVE server with Ryzen 5 1600 years ago, that’s still good enough to run dozens of LXC, GUI loaded Debian VMs and Arch virtual machines.
If anything, you need a decent amount of RAM for demanding home server tasks. But given that we’re talking about a gaming machine, your old friend probably has at least 8 GB of memory – and that’s more than enough for average tasks. For reference, I’ve been running my first gen Ryzen system with 16GB of RAM for the first couple of years, and while its 6-core, 12-thread processor sounds like a bottleneck, I’ve run into many situations where memory usage drops into the red zone, while CPU 50% will creep along nicely with us. In fact, I recently had to drop another 16GB DDR4 stick into the machine to get the 26B model working with a TN drain on a nearly 10-year-old machine. While we’re on the subject.
Even your old GPU will come in handy for certain server projects
A terrific LLM hosting can serve as a workstation
Along with your gaming PC’s superior CPU and memory provisions, you probably have a GPU stuffed into your aging rig, making it a surprisingly great addition to your makeshift workstation. On the self-hosting front, certain applications can make good use of the GPU’s processing capabilities. For example, a. if you plan to run Frigate powered Network Video Recorder you can arm it with your graphics card for AI-powered object detection on your server, and it’s significantly more powerful than a CPU-only setup when you have a bunch of security cameras around your house. Likewise, even a dinosaur GPU can handle hardware-assisted transcoding tasks in Jellyfin (or Plex if you’re willing to shell out the extra money for its premium license). Even Immich You can use the GPU for face recognition, object detection, and intelligent search tools powered by machine learning.
But these standalone tools aside, the biggest benefit of a gaming GPU is its ability to handle LLMs. And no, I’m not talking about the little 4B, 7B or 9B models either. Ever since the Mixture of Experts models started making the rounds, even older GPUs can have large LLMs, provided your gaming machine has enough RAM. Heck, I posted one Gemma-4-26B-A4B on GTX 1080 – ten year old GPU attached to my ancient first gen Ryzen rig. Given that my token generation rates are almost always above 15 t/s, you’re bound to get better performance if your graphics card is even remotely better than mine (which isn’t that much when you compare a modern mid-range card to something released ten years ago).
And that’s just LXC and containers. On the virtual machine front, you can host a remote gaming VM that pairs with the GPU to run older titles, although the virtualized nature of this setup will result in slightly lower performance than a bare-metal setup. I tend to avoid VM switching these days because I have a lot of LXCs attached to my GPU and trying to access the same graphics card from different virtual machines is not feasible in a non-SR-IOV setup like mine.
You can arm them with several PCIe cards
If your older gaming PC doesn’t have a Mini-ITX motherboard, you’ll have a bunch of spare PCIe ports in your system, even if your big GPU covers more than one. While these blank sockets are useful enough for conventional systems, their real utility becomes apparent once you start setting up your home lab. Let’s say your old gaming machine has 1G Ethernet capabilities. Arming it with a 2.5GbE network card can boost its file transfer speeds if you have a switch that supports this standard. The additional Ethernet at your disposal also allows you to set up virtualized firewall instances without having to configure VLAN-based solutions.
On the storage front, you can add extra SATA ports to your machine, throw in a few HDDs and use it as a backup drive. Alternatively, instead of relying on slow hard drives from NVMe-to-PCIe adapters, you can use these blazing fast drives as virtual guest storage pools.
With the right settings, your system will not consume much energy
The biggest complaint about reusing old systems as servers isn’t their utility—it’s their tendency to consume extra power compared to minicomputers and thin clients. To be honest, outdated computers are less energy efficient than their modern counterparts. However, there are many ways to reduce your elderly playmate’s susceptibility to electrocution.
If you’re a pixel-clocking overclocker like me, you might want to dial those amp clock speeds all the way down before arming your old rig with a home server platform. Likewise, enabling Eco-Mode, ASPM, C-states, and other power-saving options may seem like a no-brainer to FPS watchers, but these settings can go a long way to reducing your computer’s power consumption. Finally, if you’re on Proxmox, I’d recommend switching the CPU scaling manager to power saving mode, as it can massively reduce the idle wattage of your temporary home server.
