I have a QNAP TS-253D (Celeron J4125, 4GB RAM) hosting all my files. I used to have Jellyfin running on it in a Docker container, but it performed really poorly (which is expected ig). It used to take forever to stream a 1080p movie, and seeking back and forth would freeze the whole thing.
Then I moved my Jellyfin setup to my desktop PC (i9-10850k, 16GB RAM, 2080 Super), the files are still on my NAS. It performs much better now, streaming is a breeze and it almost never freezes or anything.
Problem is, it eats up all my RAM. My RAM usage is 99% almost all the time someone uses Jellyfin and it significantly hampers my regular work on my desktop. I can upgrade my RAM to 32 or 64 GB, but would that solve the problem?
If not, what is the cheapest mini PC or home setup that I can do that’d free up my desktop but still give me similar or at least good enough performance?
Thank you for your advice.
Very odd: on my computer it never uses more than 1GB of RAM. As for the limitations of your first setup, that sounds like a bottleneck when it comes to transcoding, which can occur because of a lack of GPU or CPU power, depending on whether or not you had hardware acceleration enabled. I would expect it to run much better on your new setup, but the RAM thing is worth investigating. I was running Jellyfin on 12GB of RAM until recently, and never hit 100% usage.
I am using a cheap custom setup with a Ryzen 3600 and 16 GB of ram with TrueNas and works ok. Jellyfin takes a few seconds to get started but runs fine otherwise. I can even run 4k streams but only works smoothly if i encode them with the fast decode option. This was very cheap because of the old hardware, about $300 before the drives.
The only thing I would change is to get a CPU with integrated graphics. Some can help with decoding and massively simplifies the first time setup.
I run Jellyfin on a laptop that was too slow to use for anything else and an external hard drive. Intel Celeron N3060, 2 cores, 2 threads, less than 4gb ram, 32gb eMMC storage.
I use a stack of old, retired laptops, including one from 2010, that I installed Gentoo on and setup Xen as the Dom0 hypervisor. From there, Podman and Kubernetes, and then containerized plex. I’ve never had any problems.
Running Jellyfin on a 5th gen i5 that I’ve had sitting in a closet for years doing nothing.
Running Ubuntu with 16 gb of ram. Transcodes single streams just fine. Most of my files are MP4 anyway. I run everything at 1080.
I ran for a bit on a Raspberry Pi 4. That seemed a bit strained, but OK.
I didn’t like the reliability of the Pi.
I run my truenas scale on a 17-4790 32gb ddr3 and run plex on it. Use a 1060 3gb for transcoding
I run Plex in Unraid on a 6th Gen i7 with zero perf issues and also only using the iGPU. Multiple concurrent streams no problem.
running on a single core 2 gig cloud server. just set filter to h264 so there is no transcoding.
works for me.
I’m doing something similar except now I’m running out of space on my 1 TB volume. Now I need to upgrade and/or figure out a long term solution to convert to H265. Some of my files have file size differences on the scale of 400 MB vs 2 GB.
also delete redownloadable media. most of the media i want to watch is downloaded 10 minutes prior. I just leave hard to find media always there.
I’m doing something similar except now I’m running out of space on my 1 TB volume. Now I need to upgrade and/or figure out a long term solution to convert to H265. Some of my files have file size differences on the scale of 400 MB vs 2 GB.
run tdarr. schedule it to run at downtime.
I’m using a raspberry pi 4 8 GB. It’s not a problem until someone transcodes, and that usually happens with H265 HEVC media when the person is on a browser that doesn’t support it which is most browsers these days. If the person is doing directplay (click the gear during play and click on Playback Info). If it’s forced to transcode, it will tell you why.
Direct-play in native resolution and codec should give you perfect performance and consume almost no resources since it’s essentially just a file transfer. I find performance works best when I use the Jellyfin Media Player app on my computer or the Android app.
Same here. Working very well on my Pi 4 8gb.
Same here. Even 4 different playbacks without transcoding arent a problem.
Raspberry pi 4 (4gb) with only direct stream/direct play. Works perfectly as long as the media is compatible
I bought a used computer off of Facebook marketplace for $200. It has a Xeon processor from 2014 and 2 TB of hard drive space and works like a charm. I can add 2 more 4 TB drives if I want.
Something is wrong if Jellyfin is using more than 2gb max.
I run Plex and Jellyfin and about 20 other containers in 8GB. They perform really well even though Plex doesn’t support hardware transcoding. I only have an i5 7500.
I use a beelink mini s w/ an N5095 Jasper Lake processor and 8GB RAM. Hardware acceleration supported for the vast majority of my media, and I can do 2-3 simultaneous streams without issue (I haven’t tested higher since it’s usually just my wife and I streaming).
How is your RAM being used? Look at CPU-X in the “System” tab. If the memory is mostly used for buffers and cache then it’s not a problem, you want it to be used like that.
I ran my NAS for years off an i5 (Kaby Lake) with 4 GB of RAM and 32 GB of NVMe storage and it handled any of the usual media servers just fine. I’ve used them all, Plex, Emby and now Jellyfin.
Have a look at this table, get the cheapest used Intel CPU you can find that fits your transcoding requirements, slap it on a board with enough SATA connectors and 4 GB of RAM and you should be good to go.
Docker should not have a large impact, I have 15 containers running right now and they only use 2.5 GB of RAM in total (for reals, without buffers/cache).