Plex Server Build Guide – Windows

Skill Level: Advanced

This is a guide to building a dedicated Plex server running on Windows, this is not a definitive guide. Rather a guide of recommendations that I have found useful.

Server Recommendations

Operating System: Windows Server 2016 supporting Hyper-V

CPU: Minimum 16 cores. You are going to want to find the right balance here between speed/cores/wattage, at a minimum, I would recommend no less than 16 cores.

Memory: 64GB (the more the better)

Network: Two network interfaces (1 VM, 1 Host), plus any additional interfaces needed for NAS connectivity if needed.


  • OS: 120GB (Local RAID disk, hardware not software)
  • Plex Data: 256GB SSD (Local RAID disk, hardware not software)
  • VM Disk: 512GB (NAS redundant disk or Local RAID disk, hardware not software)
  • Media Disk: as much as you can afford (Local or NAS) with the room or a plan to grow. I tier personally tier my data so that the more recent media is on a fast disk, common media is on normal disk and old uncommon media is on the slow disk.


Network Recommendations

If you plan on sharing your plex media server with your friends you are going to want a very fast internet connection with priority on the upload speed, so if your ISP has an option to obtain a faster upload you will want to look at that.

The Router

In my experience, the router your ISP will provide you will be junk. Get a router that will handle the workload (bandwidth) you want to throw at it. I recommend the Linksys WRT1900AC wireless router with DD-WRT firmware loaded on it, you may consider applying QOS rules on such a router as it makes sense.

The build

Start by setting up your Plex host on the OS disk then set up two virtual switches one external and the other internal.

At this point you have a server, network and storage, the next step is to set up some base VMs.

  • APP001 (4 core, 8GB RAM, 120GB Dynamic OS Disk, Network on both vSwitch)
    Role: This will run applications at are beneficial to the Plex media server, I recommend Radarr, Sonarr, Ombi, Tautulli and Cardigann

  • PLEX001 (16 core, 32GB RAM, 120GB Dynamic OS Disk, 200GB Plex Data Disk, Network on both vSwitch)
    Role: This will run just the Plex Media Server application
  • TOR001 (2 core, 8GB RAM, 120GB Dynamic OS Disk, Network on both vSwitch)
    Role: This will run uTorrent.

For each VM assign a LAN IP to the interface on the external virtual switch
For each VM assign a Private IP to the interface on the internal virtual switch
Configure the host file on each VM so that they communicate with each other over the internal virtual switch.

Installing Plex Media Server

  • If you don’t already have one purchase a Plex subscription
  • Download Plex Media Server,
    Note: I always recommend installing the latest version of Plex Media Server, unless the at version is like x.x.0.x
  • Install the Plex Media Server with an installation path that points to your Plex Data Disk.
  • Purchase, download and install SoftPerfect RAM Disk,
    We are going to use a RAM disk for the transcoder temporary directory so that we don’t destroy the SSD’s with writes.
  • Allocate a 20GB RAM disk with no hard drive emulation and do not save contents to an image.

Configuring Plex Media Server

Here I will provide guidance no how to configure the server behaviour, not how to organize content etc..

  • Create the following Powershell script and schedule it to run every 5 minutes

    $affinity = [int]65520
    $PlexTranscoderProcesses = Get-Process -Name PlexTranscoder
    foreach($PlexTranscoder in $PlexTranscoderProcesses){
         ($PlexTranscoder).ProcessorAffinity = $affinity

    Details: in the example script provided above on a 16 core system will limit transcoder sessions to CPU cores 5-16, thus leaving cores 1-4 always available for the Plex Media Server process itself.


  • Set BackgroundTranscodeLowPriority to 1
  • Plex Server / General
    • The path where local application data is stored = {This should be pointing to the Plex Data Disk}
  • Plex Server / Library
    • Run scanner tasks at a lower priority
  • Plex Server / Network
    • Secure connections = Required
    • Remote streams allowed per user = 2
  • Plex Server / Transcoder
    • Transcoder quality = Prefer higher speed encoding
    • Transcoder temporary directory = {RAM Disk}:\Sessions
    • Transcoder default throttle buffer = 60
    • Background transcoding x264 preset = Ultra fast
    • Maximum simultaneous video transcode = 12 (recommended for 16 core system)
  • Plex Server / DLNA
    • Enable the DLNA server (unchecked)

The Observable Internet & Darknet

The internet is a super massive exponentially expanding ice burg floating in a limitless ocean in volume and depth. Inside the ice burg is a small singularity much like a black hole that pulls information into it likely never to be see again, however the speed in which data is being pulled into the singularity is incredibly slow and as data is accessed it is pulled further away from the singularity. The observable internet which is the part of the ice burg above the water line, current estimates calculate represents at most 4% of the internet, the rest is known as the Darknet which is unknown to popular search engines. On average 99.995% of internet users will only directly interact with within the observable internet their entire life and are largely unaware the Darknet event exists. Much like when it was popular belief that the world was flat.

The majority of the users that do interact with the Darknet are comprised of intelligent agencies, large corporations and big data mining companies, both black and white hat hackers and criminal organizations. So much unorganized data exists in the Darknet that when most companies attempt a project in big data research that includes Darknet data, the often fail or retreat to working within the observable internet in favor of reduced complexity.

However there are a few creative big data architects and engineers that are able to see beyond the general disorder and chaos of the Darknet to discover insight. After many years in mining big data both in the observable and Darknet of the internet you come to accept the real challenge is not figuring out what to do with all that data but learning to develop methods to ask the right question of the data. The right question is infinitely more important than the answer, unless you don’t mind waiting forever for an answer.

To ask the write question you have to understand how data evolves within the internet:

  • If something is put online it will exist online forever, at least beyond the human lifetime.
  • If the internet is an ice burg, expansion occurs on the outer layer and internally within.
  • Users interacting with the internet are like small lightning strikes on the visible surface of the internet, it starts like a small point of data then like the branches of a tree expands out across the observable internet then down into the dark net leaving pieces of data all along the way.
  • Users leave data footprints all over the internet when they interact with services online and they may not be even be aware or accept it as the nature of online services.

There are many popular data mining designs and with the more widely available machine learning technologies we are discovering more and more data insights all the time. Once cloud computing and faster connectivity and storage becomes more widely available and adopted with the use of quantum computing answers to some of the most complex data questions will become easily accessible.