Intel NUC as Plex Media Server.

Soldato
Joined
13 Feb 2004
Posts
2,656
Location
South Shields
Sorry if this is the wrong sub forum to ask this in, the question could be asked in numerous places here.

I have a Synology 212j, a very old NAS with 4tb of storage and no transcode support via Plex media server.

My fiendish plan was to emulate a setup I had during the summer where I utilised a Mac Mini with a thunderbolt external SSD as a media server but using the Synology as an all in one solution. The Mac had since died (presumably a logic board failure, it too was very old).

We use a PlayStation 4 Pro as a media player downstairs and have an Amazon Fire Box as the media player in the bedroom, hence the requirement for transcoding.

As the Synology does not support transcoding I have been looking at different options for a server which essentially reads from the NAS. I am planning on using a SAB/SONARR/RADARR solution to organise the library for me, I am also potentially going to do most of the administration via a Remote Desktop connection on a laptop from downstairs.

All internet connections within my household, bar the obvious hand held devices, are Ethernet.

Are Intel NUCs a decent solution? I would like something with a little grunt should we decide to invest in a UHD TV in the coming future.

I don’t plan on using the Synology or the server for anything over than media playback at present.

Thank you for any feedback or advice :)
 
Rather than a NUC it may be worth looking at various SFF "mini" PCs - e.g. Lenovo Tiny, Dell Micro, HP Mini - they can often be had cheaper than NUCs and have a little more expansion option.

Keeping the Synology purely as a NAS to store Media should be fine.

Regardless of what you use for a Plex server, the ideal should be to ensure that you have a suitable client, and media in a suitable format that transcoding isn't needed.
 
Thank you for the feedback.

After reading online I’m gathering that an 8th gen onwards Intel CPU has benefits due to hardware acceleration.

Bearing in mind my relatively low user count, do you think an i3 would suffice or should I stick with an i5?

I’m also tempted by 16gb of RAM, but I’m unsure if that would be overkill?
 
Cheers randal :)

I'm actually hoping to do similar - well, hoping to finally organise my media onto one of these Lenovo Tiny boxes and set up Plex. I've used them for batch transcoding 1080p video before.

From what I've read it should be good for transcoding 1-2 4k streams but may run out of puff if there are more clients requiring transcode. Obviously as mentioned above, the ideal is to have everything stored in suitable formats to stream directly rather than transcode.
 
Depend on what you trying to transcode. Intel CPUs has integrated GPU that’s why it can cope with hardware level transcoding.

to utilise hardware transcoding you need Plex pass (paid version of Plex)

As previous suggested, keeping you synology as a data deposit and getting a power friendly system as a server/download box is a good shout. The Haswell i5 will not be able to cope with 4K transcoding tho.

https://en.m.wikipedia.org/wiki/Intel_Quick_Sync_Video
 
I used to tell people transcoding 4K was a bad idea and transcoding 4K HDR was stupid, then they fixed tone mapping, so it’s now not stupid at least, but if you must transcode 4K, doing it in hardware is the only logical way and it’s still really not best practice. When it comes to Plex, HW Transcoding requires a (paid) Plex Pass, stick to 6th gen intel+iGPU onwards, low end is fine as long as it has the 6th gen iGPU as 5th gen lacked decode support from memory and bugger all really changed after that. Also consider if for example you pipe live TV to Plex, you may have to transcode that, also any unsupported video will be done in software, along with any audio transcodes and certain types of subtitles, so you will still need a CPU that can handle some form of software transcoding ideally.

Not to tread on anyone’s toes, but unless you specifically need an ultra SFF, (great ESXi clusters btw) my advice is don’t, and that’s coming from someone with 5 of them for various jobs. Just before Christmas I grabbed a Lenovo S510 SFF i3-6100/4GB/500GB from eBay, £60 delivered on a BIN, desktop CPU, easily upgradable, cheap RAM, LP PCIe, better cooling, heck a new Pi4 4GB set-up costs more. The point being you get a lot more for a lot less if you avoid USFF’s and go SFF instead, the ‘small tax’ is quite large.

Storage wise you have two options, local (in your case the NAS) and keep buying/managing storage and paying for power/running out of space, or the cloud. The latter costs less and let’s you store unlimited data, but relies on you having a not awful internet connection and Google’s potential policy changes.

Management wise I don’t understand why you would want/need Remote Desktop? You’ll presumably be running services that have web management built in at the core like Plex, Sab/Get, Sonarr, Radarr, Tautulli etc. and automating the workflow. As you can manage everything from any device that can run a browser, why would you RDP in to then run a browser? Also what exactly are you planning on managing? Literally the only thing that you should need to touch post config (and only because you are running limited storage so automated importing is potentially a problem) is telling the *arr’s what to look for. Once you realise you can manage everything via browser, I would suggest skipping Windows, unless you have some other windows specific need you forgot to mention, Ubuntu/Debian and docker are a much better choice for this sort of usage.

The Haswell i5 will not be able to cope with 4K transcoding tho.

https://en.m.wikipedia.org/wiki/Intel_Quick_Sync_Video

Someone needs to read his own links, Haswell will transcode 4K H264 in hardware, it won’t do 4K (or any other resolution) HEVC/H265 in hardware.
 
I can chime in now I've managed to get some sleep. :)

I've been running a Generation 8 NUC (NUCi5BEH) with 32GB RAM, and a couple of SSDs (NVME + 2.5") as an ESXi server for about 18 months now. Mass storage is handled by an old Synology DS1813+ populated with 3TB Reds.

For the most part this has worked pretty well, I've got a couple of Docker VMs (one for internal facing, one for external facing services), PiHole, Plex, a Debian box for building sources and a few others for testing various things. One of the main limitations I've found with the NUC was network interfaces. Having a single NIC is very limiting in an ESXi world, so was forced to use USB NICs which is generally a cardinal sin but I can't say I've ever had stability or performance issues for the use cases I have.


Specifically for the Plex VM, I had no problem transcoding multiple 1080P streams (2 cores / 8GB RAM allocated) within the house. It's only become necessary to throw more CPU at the VM since 2020 happened as we have family and friends using the box now, so more likely to 3-4 people hitting it simultaneously.

To level with @Avalon's point, whilst the NUC is very power efficient and quiet you do pay quite a high price for that order of magnitude smaller than SFF. USFF is great, but very limiting for upgrades down the line. To that end my current project is replacing the NUC and Synology with a single ATX sized host so I can make use of more common components.
 
Do you think removing the PlayStation and replacing with a fire tv downstairs, which natively supports MKV etc, may be the better option?

Then simply run Plex server on the Synology to serve the content direct across the network!
 
Ok, GF wants to keep PS4 as the media player downstairs so back to building a server.

With reference to the management of the system, I wasn’t aware and hadn’t researched thoroughly enough the software management side. I’m happier that it can technically be an automated system though!

Is Ombi worth researching also?
 
Ok, GF wants to keep PS4 as the media player downstairs so back to building a server.

With reference to the management of the system, I wasn’t aware and hadn’t researched thoroughly enough the software management side. I’m happier that it can technically be an automated system though!

Is Ombi worth researching also?

If it's just you and the OH, do you really need to handle requests via Ombi? I mean you live with her, surely you just let her have at the *arr's and she can add what she likes or if its only the odd thing, she can talk to you :D Ombi is only really useful for remote users to request content, save it for that annoying family member who you wish you had never pinned, or better yet don't pin them ;)
 
So.. I’ve just picked up a Dell Optiplex 3050 SFF with the following specs:

i5-7500
8GB DDR4 RAM
120GB SSD
Win 10 Pro

For £170 delivered, which I’m pretty happy with tbh and I wouldn’t have got that deal without you guys!

My budget was ~£350 so I’m going to invest in 16GB of DDR4 and I’m now looking at internal storage.

I’ve been reading the specs of the motherboard and the system supports M.2 PCI-E drives.. which I’ve read are good for cache drives.

Is this something I should be thinking about? Given that I’ll be streaming over the internal network?

I’m going to buy one for the speed increase and storage increase.. just wondering if I should buy an M.2 for Cache and a standard SSD for internal storage?
 
What’s your internet connection like up/down? Cloud storage or running a hybrid cloud/local set-up could work reasonably well.

M.2 is no different to SATA in this context, NVMe is faster, but not massively so as it’s mainly small files (meta data). While 8GB is ample for Ubuntu + Plex and the general range of apps you are likely to need, it’s likely something I would just throw £30 at for 2x16GB and never have to think about again (same online marketplace you probably got the Dell from), more is always better with RAM, and transcoding to RAM saves SSD wear. 120GB is enough and if you are transcoding to RAM anyway, it’s not like wear is going to be a concern, it’s only for the OS+apps and meta data, the only concern would be if you download a big file using news for example, let’s say 50GB, then it needs potentially another 50GB to unpack/process/repair, obviously that could cause space problems. If you have another SSD kicking about, great, if not an old mechanical drive will work and it can be slower to unpack etc. and won’t slow down your main system.

You can also manage it via app, LunaSea on iOS for example makes adding content easy if you don’t have a laptop etc. to hand.
 
Last edited:
What’s your internet connection like up/down? Cloud storage or running a hybrid cloud/local set-up could work reasonably well.

M.2 is no different to SATA in this context, NVMe is faster, but not massively so as it’s mainly small files (meta data). While 8GB is ample for Ubuntu + Plex and the general range of apps you are likely to need, it’s likely something I would just throw £30 at for 2x16GB and never have to think about again (same online marketplace you probably got the Dell from), more is always better with RAM, and transcoding to RAM saves SSD wear. 120GB is enough and if you are transcoding to RAM anyway, it’s not like wear is going to be a concern, it’s only for the OS+apps and meta data, the only concern would be if you download a big file using news for example, let’s say 50GB, then it needs potentially another 50GB to unpack/process/repair, obviously that could cause space problems. If you have another SSD kicking about, great, if not an old mechanical drive will work and it can be slower to unpack etc. and won’t slow down your main system.

You can also manage it via app, LunaSea on iOS for example makes adding content easy if you don’t have a laptop etc. to hand.

To be totally honest I’ll probably stick with windows as the OS (I’m not very experienced with Linux based environments) so will likely up to 2x16gb RAM, may well purchase a 500gb SSD/NVME too as I’m a little sceptical of a 128gb drive being big enough.

I’ll likely setup the system to download to the SSD then unpack to the NAS. Unless I can download directly to the NAS.

My internet connection is ok, 350mb Virgin fibre :)

How difficult is Linux? And how much of a performance improvement would I see using Linux rather than windows?
 
Go with whatever you feel most comfortable with, but Windows and hardware transcoding doesn’t always scale well/reliably. I tend to favour what’s easiest to run/maintain, and gives the best performance/£ advantage, having used a lot of different platforms both locally and remotely, this is easily a win for an underlying BSD/Linux OS and docker. Installation is stupidly simple with something like PlexGuide or CluodBox or one of the AIO docker solutions that are being developed. Managing the apps discussed is the same regardless of the underlying OS.

As to 120GB, a lot of the PG user base have VPS’ around that mark, they tend to run larger libraries than you will be and with faster connections, but I agree it’s easier (read you can be less careful) with a larger drive. You could always have a go with cloud storage and see how you get on, once you can pull media faster than you need to play it from the cloud you start to see storage in a different way.
 
A little update on this project.

So.. I’ve now got the 3050 and after a few initial problems (CMOS and RAM related) I’m now up and running from a hardware point of view.

I ended up with a 500gb 970 Plus NVME for an OS drive on the machine, removing the 120gb SSD (now installed in a 10 year old laptop which incidentally feels waaay quicker) and I’ve maxed out the board memory at 32gb DDR4 due to memory prices being so low for slow frequency RAM.

My plan is to allocated some of this 32gb DDR4 to a dedicated RAM drive for media caching. Is this generally a good idea? I’ve only read good things about this approach. How much should
I allocate, 8GB?

Next thing I need to figure out is.. where and how do I run SAB/SONARR/RADARR.

Having watched a few videos on YT, I’m thinking that I run these three packages within Docker directly on the DS212J which will house my media files. Is this perceived to be the correct approach?

From a logical POV it makes sense, as DSM can handle all of the downloading, renaming, organising of files whilst Plex Media Server on the Windows machine can solely focus on delivery of media.

Does this make sense? Am I missing something or making any rookie mistakes?

Thanks for all your help, apologies to those whose advice I haven’t followed. I haven’t had a tech project in years and had some extra cash laying around so it’s been unusually easy for me to pull the trigger on OTT stuff.
 
OK take everything you just said and do the opposite, literally.

NVMe is more than capable of handling meta data from way more clients than you’ll ever need while coping with transcodes and unpack/repair/move operations, you can transcode either to RAM (but expect zero performance gains vs NVMe) or just leave it to use the NVMe drive - you won’t be saturating IO on it unless you have 10Gbit WAN.

I dread to think what Sonarr/Radarr and SAB would do to a 212j, it’s a comparatively ancient single core ARM chip with 256MB RAM and also has to run the OS and whatever else you have running, it’s a bad idea, especially with a reasonably fast connection.

I run Plex and most other things in docker, they’re easier to maintain, update and manage that way, the idea of running them in Windows leaves me cold, but you’ll need to install them as normal bare metal, set them up in the normal way via the web interface and get them talking to the download client via API key and Plex to notify it of changes. Once everything is processed it gets moved to the shared folder on the NAS. If you’re using HW transcoding it’s not like the CPU won’t just be sitting idle most of the time anyway.
 
OK take everything you just said and do the opposite, literally.

NVMe is more than capable of handling meta data from way more clients than you’ll ever need while coping with transcodes and unpack/repair/move operations, you can transcode either to RAM (but expect zero performance gains vs NVMe) or just leave it to use the NVMe drive - you won’t be saturating IO on it unless you have 10Gbit WAN.

I dread to think what Sonarr/Radarr and SAB would do to a 212j, it’s a comparatively ancient single core ARM chip with 256MB RAM and also has to run the OS and whatever else you have running, it’s a bad idea, especially with a reasonably fast connection.

I run Plex and most other things in docker, they’re easier to maintain, update and manage that way, the idea of running them in Windows leaves me cold, but you’ll need to install them as normal bare metal, set them up in the normal way via the web interface and get them talking to the download client via API key and Plex to notify it of changes. Once everything is processed it gets moved to the shared folder on the NAS. If you’re using HW transcoding it’s not like the CPU won’t just be sitting idle most of the time anyway.

The way you started this reply was amazing, I’m a rookie at all this and need all the help I can get.

Whilst I can go against the grain and buy hardware which is more powerful than I need, I definitely cannot ignore configuration advice!

What I was trying to avoid was over using the NVME drive, thinking that if I download to the drive, process files on the drive and transfer from that drive to the NAS it would be essentially wearing the drive down overtime.

I do have a ~8 year old Samsung EVO 1TB SATA SSD sitting around, would it be better to download, unpack etc onto that drive to transfer to the NAS?

I’ve heard a lot about Docker, I’ll look into it but do you have any links or advice? Is it a separate OS? How hard is it for a novice to work on (I have extremely low amounts of experience with OS bar Windows and Mac OS).

Really appreciate your help!
 
The 970 Evo Plus is rated for 0.3 DWPD, or in other words 1/3 of its capacity can be written per day, every day for 5 years before it reaches the threshold Samsung rates it to, it isn’t dead at that point, just out of warranty coverage. You have a 2 bay NAS, where exactly are you planning on putting the 300TB of data?

Also consider that the published endurance is the minimum you can expect, this illustrates what can happen: https://techreport.com/review/27909/the-ssd-endurance-experiment-theyre-all-dead/

Either way, write endurance isn’t something I would worry about. The 1TB Samsung would have saved you buying an NVMe drive at all - the real world differences in performance is negligible and it’s not like you’ll have 15 concurrent users. That said if you want to use a RAM drive and transcode to that, it won't do any harm - just make sure it's big enough for however many streams you need to transcode concurrently.

While docker is available for Windows, I wouldn’t suggest it’s going to make your life easier under Windows, it’ll just add complexity that you probably don’t really want or need (or you wouldn’t have chosen Windows).
 
Last edited:
Back
Top Bottom