For those who’ve been fortunate sufficient to head via a PC improve or two in the previous few years, there is a just right likelihood you have got had a GPU leftover. It is almost certainly sitting in an anti-static bag in a drawer someplace, completely dormant, or you may have been that means to place it up on a pre-owned market however simply have not. Whilst it will not be adept at operating the most recent titles, previous gaming GPUs have a ton of doable in a house server construct, the place they may be able to excel at native media transcoding, GPU passthrough for VMs, and native AI.
The most productive worth GPU for Plex transcoding is far other (and older) than you’ll be expecting
You could have already got one
Transcoding is the place native GPUs will also be most precious
A large spice up for high-resolution media streaming
Essentially the most instant and common case for an previous gaming GPU in a house server is {hardware} transcoding. If you are operating Jellyfin or Plex to flow media to gadgets round your own home or over the web, your server is sort of for sure transcoding movies. Transcoding occurs when recordsdata wish to be transformed from one decision and bitrate to every other so as to fit what the streaming instrument can take care of.
By way of default, that paintings falls solely at the CPU, and it is normally moderately pricey, even supposing there is an exception to this which I will quilt later. A unmarried 1080p transcode can pin a couple of CPU cores for a longer duration, and if a couple of persons are streaming concurrently, the bottleneck turns into very actual, very speedy.
An previous gaming GPU will be capable to harness a {hardware} encoder to take at the brunt of the encoding paintings. Nvidia makes use of NVENC, and AMD makes use of VCE, however NVENC is the most popular means. When Jellyfin or Plex offloads transcoding to this {hardware}, the CPU slightly notices the workload is occurring in any respect. One thing as modest as a Pascal-era GTX 1050 can take care of a couple of transcoding streams without delay.
I became a lifeless GPU right into a {hardware} encoder, and it is best possible for my NAS
Waste now not need now not
GPU passthrough is the place issues get actually thrilling
If you are operating Proxmox, a devoted GPU is a useful gizmo
Transcoding is only one use case for that previous GPU, however despite the fact that you are now not operating a media server, you’ll be able to nonetheless put it to nice use. For instance, in a Proxmox setting, the usage of that gaming card for GPU passthrough shall we a digital system declare it as though it had been its personal. That suggests a Home windows gaming VM will have an actual graphics card for show output and light-weight gaming, or a devoted container can run AI inference workloads with out touching the host gadget. Native huge language fashions, symbol technology with Solid Diffusion, and audio transcription with Whisper all receive advantages significantly from GPU acceleration. Your previous card almost certainly does not have massive quantities of VRAM to play with, so you can be the usage of smaller fashions and lighter inference duties, however it is completely nonetheless price throwing it into your gadget.
I constructed my first house server for only $200, and it changed 4 per month subscriptions
An previous, dormant gadget has the prospective to interchange subscriptions price masses of bucks a 12 months
Previous GPUs do have their boundaries
Constrained through their nature
Those previous gaming GPUs do display their age, and they are now not a silver bullet for giving your own home server extra software. AMD playing cards particularly have had a traditionally inconsistent open-source motive force tale on older generations, which may make setup on headless servers extra sophisticated than it will have to be. NVIDIA’s proprietary Linux drivers are extra predictable, however they are now not proof against version-compatibility problems, and passthrough with each playing cards will also be fickle. In case your motherboard and CPU do not toughen IOMMU, you are mainly out of success, or even then, it is not simple.
Energy intake is price factoring in too. In case your card was once a former flagship, having it run 24/7 in your house server construct may now not make a ton of sense if you are looking to stay energy prices down.
6 GPUs which can be formally too previous in 2025
Those GPUs can now not deliver it in fashionable titles
It isn’t doing the rest laying dormant
If in case you have a unfastened PCIe slot, tossing it in prices nearly not anything
Loose upgrades are the most productive more or less improve, and when your previous gaming GPU is sitting in a table drawer or its field in a closet, it is not incomes its stay. For transcoding in particular, the NVIDIA plus Jellyfin mixture in Docker is one of the documented GPU configuration within the self-hosted neighborhood. It isn’t plug-and-play, however it is shut, and the payoff is straight away measurable if you are coming from tool transcoding. The similar is going for compute. Troubleshooting GPU passthrough is worth it if you are frequently spinning up VMs that will get pleasure from GPU acceleration.
24GB VRAM will have to be the brand new minimal for many GPUs
An previous gaming GPU is the perfect house server improve round
That GPU within the drawer is not going to come back again into your gaming rig anytime quickly, and it is almost certainly now not going to fetch a significant worth at the secondhand marketplace. What it might probably do is select up a 2nd existence doing paintings it is nonetheless actually just right at. Taking up transcoding tasks, powering VMs with an actual show output, or changing a couple of small cloud products and services with inference that is if truth be told helpful are all issues previous gaming GPUs can excel at.
