Repurposing older Xeon hardware for a Jellyfin build – Is the power-to-performance ra

[dapemo9713]

Hey everyone, I’ve been a long-time lurker in the self-hosted community, and I’m finally ready to make the jump over to Jellyfin. I’ve spent the last couple of years using a basic NAS for file storage, but I really want a proper media server experience without the licensing headaches or "home-phoning" issues found in other platforms.
I’m currently at a bit of a crossroads regarding my hardware. Instead of buying a brand-new, low-power N100 box, I’m looking at repurposing an older workstation I have in my closet. It’s running a Xeon 6-core processor clocked at 1.9GHz with a 7.2GT-QPI bus speed. My personal insight is that I’ve always preferred the build quality and stability of enterprise-grade gear; there’s just something satisfying about giving a "retired" server a second life as a home entertainment hub.
However, I’m trying to be realistic about my expectations. One specific point I keep seeing in the Jellyfin documentation is the heavy recommendation for Intel QuickSync for hardware acceleration. Since this Xeon is an older architecture without integrated graphics, I know I’d be relying entirely on software transcoding for any clients that can’t direct play my library.
At 1.9GHz, I’m a little concerned about the per-core performance when it comes to 1080p HEVC files. I know the 7.2GT-QPI (QuickPath Interconnect) provides great bandwidth for moving data between the CPU and the rest of the system, but I’m worried that raw brute-force CPU transcoding might turn my server room into a sauna. I’m mostly looking to serve about 3 or 4 users simultaneously, primarily on local 10-bit color depth files.
I’ve considered throwing a cheap, low-profile GPU in there to handle the transcodes, but I’m trying to keep the power draw somewhat reasonable. I’d love to hear from anyone who is still running Jellyfin on older multi-core Xeon setups. Does the extra cache and thread count of a 6-core chip help balance out the lower clock speed, or am I just setting myself up for a lot of buffering circles and a high electricity bill?
Is it still viable to use this kind of legacy multithreaded hardware for a modern media stack, or have we reached a point where dedicated hardware encoders are basically a requirement for a smooth multi-user experience?

[TheDreadPirate]

TLDR; Don't.

For CPU transcoding, that old geezer cannot keep up. If you are going to GPU trancode, you are just wasting energy powering such an old and inefficient CPU.

Regardless of the CPU you have, GPUs are pretty much a must have for a smooth experience when transcoding. The iGPU in the N100 has a capable dedicated encoder and should meet your transcoding needs.

[TheDMV]

You could run some FFMPEG benchmarks and see how badly trying to simultaneously transcode 4 files would be.

You could spend $X on a GPU and still struggle.

Or, consider separate units; your Xeon as a NAS for storage and reliability, and another unit (N100 as DreadPirate suggested) as the front end.

Or, even better, an old second-hand Apple M1 Mini. These tiny little boxes have more power and speed for basically no power draw and no noise, than any previous intel chip.

https://github.com/Digital-Shane/jellyfin-on-macos

"a base-model M1 Mac mini for just $300, easily handling eight simultaneous live transcode sessions without issue. After upgrading to an M4 Pro Mac mini, transcoding performance became essentially limitless for practical purposes."