Looking to upgrade my server

Hello all,

I’m looking to upgrade my server and was hoping to get some recommendations. I’ve tried to include as much information as possible, and any feedback would be appreciated.

Budget

£400

Current server:

Laptop with 8GB RAM, an Intel® Core™ i7-4710HQ CPU @ 2.50GHz running Ubuntu server 18.04

Requirements:

PleX server. Currently only local streams, no 4K content, streaming to an Android TV and Android phones (2 streams at a time max). I would like for a new setup to be able to be able to handle remote streams, 4/5 1080p streams, not bothered about 4k. I’ll also be introducing a HTPC into the house as well.

Side-projects. I’m a Dev so, on occasion, I’ll decide to start a side project (website/bot etc) and host it on my server. Nothing too resource-intensive, as most are just to play, if any of them got THAT popular I’d move it to the cloud.

OS: Linux.

Power: I have tried to stay away from rack servers because I don’t want something that I’ll pay through the nose in electricity for. My current server (laptop) is about 60w but I don’t mind exceeding that for a superior system.

Noise: it’s going to be in my office, which is the room next to my bedroom, so preferably something that’s not really loud.

Form factor: not that bothered: normal desktop size is fine.

Desktop size is nice, aim for low rpm 120mm or larger fans, whisper quiet. The lower power cpu and components, the less fan speed you need.

Anything Intel 7xxx gen and up using igpu are transcode monsters, can handle hevc10bit. I use a i3-8100 and its a beast for plex.

I use microATX mainboard, and a nvme m.2 to preserve all sata channels. Im sitting at 5/6sata channels used for plex storage.

Its gonna use more power than a laptop, especially if you start adding 3.5in storage drives.

Thanks for the reply!

I use a i3-8100 and its a beast for plex.

I assume this is because the 8th generation CPU’s have decent QuickSync, which PleX can utilise?

Correct. Starting with 7xxx series igpu is phenomenal.

Direct streams require little cpu, can use as little as a raspberrypi. Just passing through file access. So can your laptop.

But transcode streams… that gets into beefy territory. The 8100’s igpu is very good at video, can handle 6+ 1080P transcodes.

So when you say

4/5 1080p streams

Direct or transcode… and remember if you have friends/family whether they have the player and bandwidth they’ll f it up and transcode without knowing it. I have quite a few who have fiber connections, shieldtv, chromecast, modern LG and Sony, and plod away all day transcoding. When I first started with a macmini sandy bridge, before hw transcode. Had maybe 4 users, close family. Still was having to constantly text them, you are using up all the cpu, staaahp!. Fix your client settings.
Now I tell them once, and leave them to their own suffering.

Direct streams require little cpu, can use as little as a raspberrypi. Just passing through file access. So can your laptop.

I have another thread going, as I am experiencing performance issues - but this happens even when direct playing, so now I am sceptical about whether to even bother upgrading my server.

But transcode streams… that gets into beefy territory. The 8100’s igpu is very good at video, can handle 6+ 1080P transcodes.

Wow, 6 1080p transcodes would be more than enough. So, armed with that knowledge, I could look to get a Dell Optiplex 3060 with an i3-8100 and that should satisfy my requirements easily?

Direct or transcode… and remember if you have friends/family whether they have the player and bandwidth they’ll f it up and transcode without knowing it. I have quite a few who have fiber connections, shieldtv, chromecast, modern LG and Sony, and plod away all day transcoding. When I first started with a macmini sandy bridge, before hw transcode. Had maybe 4 users, close family. Still was having to constantly text them, you are using up all the cpu, staaahp!. Fix your client settings.

I don’t have any remote streamers, at the moment, but I’m curious what you mean by fixing their client settings? Is there something that should be set on the client-side?

OptiPlex will have the grunt. I rolled out a bunch of Precision 3630’s with i3-9100’s and the 3630 chassis has lots of drive bays to roll in local storage.

…so digging in I see several alarms.

I’m curious what you mean by fixing their client settings? Is there something that should be set on the client-side?

YES. If a client is not set to highest/maximium/directplay, it can attempt to transcode. Each client must be individually configured, on the client device. Direct streams require little cpu, but as soon as a client requests to transcode, it can wreak havok on an underequipped cpu (like your 4xxx).
I would love an option in plex, server-side, to block transcoding from the server, if nothing else to prevent errors dragging server down. It would be nice… to have on a per library basis, @ChuckPa, to apply to a 4K HDR, library, but not a 720/1080P library :wink: nudge

@sim
Going back to this thread/post you linked:

ChuckPA explains all.
Your media is in HEVC, your client can’t process HEVC, your cpu is being dragged through the mud trasncoding HEVC in realtime and has no gpu assistance.

no cost:

  1. insure your media is in a format that your clients can direct stream.
  2. insure your clients are set to direct stream/play, maximum quality.
  3. avoid PGS subtitles as that can force a transcode

low cost:
Super new clients. AppleTV 4/5 with enhanced player supports virtually all formats and can handle PGS subtitles without transcodes, but that’s exception and not the rule.

more cost:
Intel cpu 7xxx and newer can transcode on gpu all the popular formats HEVC10 etc on the igpu without dragging cpu into the mud.

Thanks again for the reply!

If a client is not set to highest/maximium/directplay, it can attempt to transcode.

I assume by this you mean the Video Quality setting? E.g. the below from the web player

image

Each client must be individually configured, on the client device.

It looks like this was already set on my main device (Sony Bravia Android TV), but updated my PMP and Android phone.

Your media is in HEVC, your client can’t process HEVC, your cpu is being dragged through the mud trasncoding HEVC in realtime and has no gpu assistance.

This makes perfect sense, for the files that are HEVC. But I am also seeing this sort of performance hit for files that Direct Play.

Honestly, I am perfectly happy to spend the money buying a new device to use as a server - I just didn’t want to waste the money if it wasn’t a CPU thing and I was just going to end up with the same problems.

Intel cpu 7xxx and newer can transcode on gpu all the popular formats HEVC10 etc on the igpu without dragging cpu into the mud.

Is there MUCH benefit to going for a 8xxx series over a 7xxx series? I took the following from here, both seem to support HEVC :man_shrugging:

Version 7 (Ice Lake)
The Ice Lake (microarchitecture) adds VP9 4:4:4 decoding, VP9 encoding (up to 10-bit and 4:4:4), HEVC 4:2:2 and 4:4:4 decoding and encoding,[14] HDR10 Tone Mapping[15] and Open Source Media Shaders.[16] HEVC hardware encoding quality has also been improved.[17]

Version 8 (Tiger Lake)
The Tiger Lake (microarchitecture) adds VP9 12-bit & 12-bit 4:4:4 hardware decoding and HEVC 12-bit 4:2:0, 4:2:2 and 4:4:4 hardware decoding[18]

I have an i5-7400 in one system. It seems every bit as fast and capable as the i3-8100 in another. That’s subjective, I’ve not done direct to direct.

7xxx has HD Graphics 630, while the 8xxx has UHD Graphics 630. Not sure the difference as even the 7xxx is supposed to handle 4k(ive not tested 4k).

The -7xxx processor is Kaby Lake, not Ice Lake.
The -8xxx processor is not Tiger Lake, the Refreshed KabyLake

If I may caution, please be very careful not to confuse the ASIC technology with the CPU designator.

-6xxx = Sky Lake
-7xxx = Kaby Lake
-8xxx = Coffee Lake
-9xxx = Cannon Lake
-10xxx = Ice Lake
Tiger Lake will be 11th Gen.

Thank you. Is there a minimum generation that I should be looking at getting? Is the 8th generation that much better than the 7th etc?

-8xxx and -9xxx are both CoffeeLake. Intel gone a bit loopy.


-7xxx can handle a lot. Each generation gets a little more performance, but no major new features.
Later gen adds nothing that you’ll use, it can hw decode VP9(10bit).
chart here:


The ‘community’ isn’t even really embracing VP9, most are drooling for AV1 to solidify.

Thanks @Menel, I thought that @ChuckPa mentioned that Wikipedia link referred to ASIC technology and not the CPU’s I would be looking at with this comment?

Hi, the Delll Optiplex looks good. How many HDD are you looking to put in it? Will you looking to have OS on SSD and Media on HDD’s mirrored or RAID in same box?

@Menel thanks for all your help during this post. I upgraded to a Dell Optiplex 7070 micro with a 9th gen i5 and converted my Laptop (which was the server) into a HTPC.

Everything is working great now!

Not sure why @ChuckPa just stopped replying in the other post, so thank you for stepping up and helping me get this sorted :slight_smile:

Saw your post lately.
Im looking at a Dell optiplex 3070(micro) i5 9500T, which has UHD 630 IGPU.
Anyone around here that thinks it’s a good PMS?
No 4K, 95% = 264x
15/20 users, 7/8 maximum simultaneously.

Like to hear your thoughts! Or anyone else!

Im looking at a Dell optiplex 3070(micro) i5 9500T, which has UHD 630 IGPU.

The i5 9500T does have Quick Sync and has a benchmark score of 8200. I don’t know how much extra oomph QS gives a CPU, but this article talks about what raw CPU power you may need.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.