Your Plex server, running on a machine in a data center

Will it keep up with your viewers?

It will if the processor can handle the conversions happening at the same time. When a phone, TV or browser can play a file directly (Plex calls it Direct Play), the server barely works. When it cannot, because of the format, burned-in subtitles or a slow connection, Plex transcodes, and that is where the processor earns its keep.

Plex gives a rough benchmark for one conversion at a time, measured in PassMark points:

• a 720p H.264 file at 4 Mbps: about 1,500
• a 1080p H.264 file at 10 Mbps: about 2,000
• a 4K SDR HEVC file at 40 Mbps, converted to 1080p: about 12,000
• a 4K HDR 10-bit HEVC file at 50 Mbps, converted to 1080p: about 17,000

Multiply by how many people convert at once. Plex's own example is four 1080p viewers needing roughly 8,000. It also says remote viewers very often need a transcode, because their connection rarely takes the full-quality file, so a server outside your home leans on this table more than one in your living room.

Memory is the easy part: Plex reckons 4 GB is typically plenty. Storage needs more thought. Besides the films themselves, Plex keeps its own database, artwork and preview files, and writes temporary files while it converts video, so an SSD for both makes browsing feel quick.

Why you will want a Plex Pass

A hosted Plex server is never on your home network, so Plex counts every video you play from it as remote playback, and since 29 April 2025 playing video remotely has had new requirements. On its updated apps, including the web app, iPhone and Android, smart TVs, Roku and games consoles, that requires one of two things: the owner holding a Plex Pass, or the viewer holding a Remote Watch Pass or Plex Pass of their own. The owner's Plex Pass is the simple route: it covers everyone you share your libraries with. Music in Plexamp and photos stay free.

You also need a Plex Pass to switch on hardware transcoding. If you would rather not pay a subscription at all, the free, open-source option is Jellyfin, which we cover on its own page.

Be aware of Plex's own position: Plex Media Server is licensed for personal use and meant to be run at home, and on rented hardware at a hosting company, Plex says, it might not work with every host. Try it with your own devices before you pay for months ahead.

Hardware that converts video for you

Hardware transcoding hands the conversion to a dedicated video engine, so more people can watch at once and 4K plays more smoothly. Plex uses two kinds on Linux: Intel Quick Sync, built into some Intel processors, and the video engines on NVIDIA graphics cards (driver 450.66 or newer). It does not work inside a virtual machine, Plex explains, because the host does not hand the video hardware to the guest. That is the plain reason a whole physical server suits Plex better than a cloud VM.

• Quick Sync: of the processors we rent, four have Intel graphics on the chip: the Xeon E3-1284L in its v3 and v4 versions, the E3-1585L v5 and the Xeon E-2286M. Each is listed with its specification on our Intel servers page. Plex accepts Quick Sync from 2011's Sandy Bridge onward and recommends Broadwell or newer.
• NVIDIA: our desktop machine in Bucharest has a Core i5-9500 and a GeForce RTX 3050 6GB of its own, and the GPU servers page lists every card we can fit to a bigger server. Plex warns that GeForce cards can limit how many conversions run at once, a restriction in NVIDIA's driver.
• Quality: Plex notes that hardware conversions can look slightly softer than software ones, mostly below 720p, and that older chips give weaker results than newer ones.

The server that fits your library

Start from where your films will live. A modest collection fits the built-in SSD or NVMe of a Quick Sync machine, which keeps conversions cheap. A collection measured in terabytes wants many drives, and the storage servers and larger rack machines take them with hardware RAID. Those bigger processors have no Quick Sync, so there you either rely on Direct Play, size the processor by Plex's PassMark figures, or add an NVIDIA card.

Whichever you choose from our dedicated servers, the machine is yours alone with root access, and nothing meters its port, so there is no allowance to run over however much your family streams. A server already in a rack is handed over within about 30 minutes of payment approval. Adding drives or a graphics card means a build, which takes 4 to 24 hours when we hold the parts and longer when they have to be bought in. There is no contract; you stay month to month.

Up and running in an evening

Plex's official Docker image, plexinc/pms-docker, is a tidy way to run it. This walkthrough assumes Ubuntu with Docker already installed, and a Quick Sync machine.

1. Turn on the firewall with only SSH allowed, then check the chip's graphics are visible: the output of the first block should include Kernel driver in use: i915. If the first command finds nothing, the last line lists display devices of every kind.
2. Sign in at plex.tv/claim and copy your claim token. Plex's own chart in the pms-docker repository says it is valid for only a few minutes, so start the container straight away.
3. Run the second block. Host networking keeps Plex's ports under your firewall rules; --device=/dev/dri hands Quick Sync to the container. The token links the server to your account on its first start and is ignored afterwards.
4. Once the server shows up in your Plex account, open port 32400 with the third block, and if the automatic check under Settings > Server > Remote Access fails, choose the option to set the public port by hand and type 32400.
5. In Settings > Server > Transcoder, show advanced settings and switch on hardware acceleration.

sudo ufw allow OpenSSH
sudo ufw enable
lspci -v -s $(lspci | grep VGA | cut -d" " -f 1)
lspci -k | grep -A3 -Ei 'vga|display'

docker run \
-d \
--name plex \
--network=host \
-e TZ="Etc/UTC" \
-e PLEX_CLAIM="claim-xxxxxxxxxxxxxxxxxxxx" \
-v /srv/plex/config:/config \
-v /srv/plex/transcode:/transcode \
-v /srv/media:/data \
--device=/dev/dri:/dev/dri \
plexinc/pms-docker

sudo ufw allow 32400/tcp

The claim token sits in your shell history and in the container's settings, so treat it like a password. Missed the window? Skip the token and reach the setup screen through an SSH tunnel instead (ssh -L 8888:127.0.0.1:32400 youruser@your.server.ip, with the login you use for SSH, then open http://127.0.0.1:8888/web). Plex's installation guide sets up this tunnel so that, to the new server, your browser looks like a device on its own local network.

Keep away from Docker's bridge mode when the server faces the internet directly. Its -p mappings get past ufw, and Plex's own bridge example publishes discovery and DLNA ports that its firewall article strongly advises a data-center server never expose. Port 32400 over TCP is the only one Plex needs open.

Whose films are these?

Yours. Plex is built for a personal collection: films and shows you hold the rights to, home videos, your music and photos. You are responsible for the rights to everything you put on the server and for everyone you invite to watch it.

Questions people ask

Is a Plex Pass required when the server is hosted?

For video, effectively yes. A hosted server is always on a different network from your TV or phone, and on Plex's updated apps remote video playback needs either the owner's Plex Pass or, for each viewer, a Remote Watch Pass or a Plex Pass on their own account. The owner's Plex Pass also unlocks hardware transcoding.

How much memory should a Plex server have?

Plex's guidance is that 4 GB usually covers it, and that Linux servers often manage on even less. Spend the rest of your budget on the processor or graphics card, which decides how many people can watch converted video at once, and on disks for the library.

Can I use an NVIDIA card in Docker?

The hardware section of the official image's README gives a container recipe for Intel Quick Sync only; for NVIDIA it just says to install NVIDIA's Linux drivers on the host. Plex's hardware acceleration guide asks for driver 450.66 or newer, so the plainest documented route is the regular Plex package installed on the server itself.

Which port does Plex use?

TCP 32400 is the one Plex requires. The other ports it lists are for discovery and DLNA on a home network, and Plex strongly recommends against exposing them on any server reachable straight from the internet, which a data-center server is.

Does Plex mind running at a hosting company?

Its licence is for personal use, and Plex writes that it expects servers to sit in people's homes, adding that some hosting providers may not work with it. Read its terms, test playback from your own devices, and only then commit to a longer billing term.