+- Jellyfin Forum (https://forum.jellyfin.org)
+-- Forum: Support (https://forum.jellyfin.org/f-support)
+--- Forum: General Questions (https://forum.jellyfin.org/f-general-questions)
+--- Thread: How to benchmark VPS for Jellyfin (/t-how-to-benchmark-vps-for-jellyfin)
How to benchmark VPS for Jellyfin - tomate - 2025-05-19
Heya,
I'm running a Jellyfin instance for family and a couple of friends. Recently my VPS (4 cores, 8 GB Ram) maxxed out when several streams had to be transcoeded (VPS is running *arr stack as well). But I can move this to a server with more storage and better connection - but low cpu/ram specs.
I've got access to several cheap-ish options for VPS, ranging from 10 Arm cores or 4 dedicated cores or 8 x86-64 cores and so on. What would be a good way to find the best performing server for the job?
I know of Jellybench but it isn't stable yet (I could work around the issue that it's downloading x86 binaries for Arm, but it still crashes on both Arm and x86-64) and it's predecessor hwatest doens't work as well.
I am not looking for solutions to get these things running, but for other sensible ways to benchmark servers for Jellyfin (and mainly CPU based Transcoding, or will a VPS be able to support hardware acceleration?)
TIA!
RE: How to benchmark VPS for Jellyfin - bitmap - 2025-05-19
You could get HWA out of a VPS that supports VA-API or QSV, which includes a broad range of Intel chips as well as any Arc card you want to throw at it. For CPU encodes, you're only looking at a few simultaneous encodes even in the newest generation unless you go with top of the line or tons of cycles. That's a big portion of the reason why HWA exists for real-time transcoding. CPU presents a major bottleneck for simulatneous transcodes and, in my experience, you get diminishing returns (i.e., the relationship between pegging all cores at a certain multiplier and adding an additional encode is NOT linear).