First NAS build - Supermicro-X11SSL-CF

[Yorick]

Yes, hw transcode means to use the GPU to accelerate the transcode operation.

If you have an RPi, that’s a static (as in not mobile) display, and that means it’s either 1080p or 4K HDR capable, and that in turn means Plex won’t have to transcode video at all. The Plex server will just send the video data as-is, and all the Plex client does is take the H.264 / H.265 content, decode that, and display it. Which means the Plex client does need enough power to handle the decode.

Transcode comes in when you’re streaming to a mobile device. 4K HDR or 1080p to 720p or even 480p, depending on receiving device. Ditto for audio, usually down to a stereo stream. If streaming to a phone or tablet isn’t a thing, the ability to transcode is not that important.

 

[ullbeking]

Thank you for sticking with me on this @Yorick... I'm not trying to convince you that you're wrong, on the other hand I don't believe I'm wrong either. The whole point of a homelab is that we can architect and construct system how we want to do them. Not like how our employer or a client wants them done. If this wasn't fun or educational I wouldn't be doing it. In fact, I'd probaly be looking to change to a completely different career.

If you have an RPi, that’s a static (as in not mobile) display, and that means it’s either 1080p or 4K HDR capable, and that in turn means Plex won’t have to transcode video at all. The Plex server will just send the video data as-is, and all the Plex client does is take the H.264 / H.265 content, decode that, and display it. Which means the Plex client does need enough power to handle the decode.

Is an RPi powerful enough to transcode 4K on the fly? Personally I'm more an Orange Pi kinda guy myself, but whatever. It's the same principles.

I explained my intented architecture in the thread I started over there -> https://forums.freenas.org/index.ph...as-as-a-transcoding-server.72728/#post-509701 You are a contributor to that thread too

I have no intention of using Plex. I would rather use VLC, FFmpeg, and maybe Kodi, maybe Jellyfin, etc. Highly preferable to use libre and open source tools. My first job, after it's set up, will be ripping a lot of music and DVD's for archives and storing in the media server (not the general NAS, which will be provisioned for serving files and nothing else) using a Xiph.org codec. The initial ripping is going to be very CPU-intensive. I have a lot of CPU power available to me.

We are not data hoarders. We don't watch TV and I've never seen The Wire, Breaking Bad, or Game of Thrones. We love particular movies and have those ones on DVD. We don't have Netflix. We have Spotify because we love music. But we listen to most of our music on a mechanical drive turntable and the music signal never even touches an ADC. We have many DVD box sets for our daughter. We're going to get a projector to watch much the digitally stored media on, e.g., movies.

Transcode comes in when you’re streaming to a mobile device. 4K HDR or 1080p to 720p or even 480p, depending on receiving device. Ditto for audio, usually down to a stereo stream. If streaming to a phone or tablet isn’t a thing, the ability to transcode is not that important.

We do listen stream to mobile devices a lot.

 

[Yorick]

Oh 100% agreed on the ability to do things your way. I'm just adding my two cents on how I'm seeing it. Infinite Diversity in Infinite Combinations and all that.

For the Raspberry Pi as a device to transcode 4k, my gut check is "nope". A quick Google also says "nope", https://forums.plex.tv/t/4k-and-raspberry-pi/235407 . Yeah that's Plex, but the needed CPU power doesn't change with another solution. Plex actually uses ffmpeg, as far as I know, so data for Plex are a good stand-in for what you'd need if you use ffmpeg directly.

My understanding is that a 1080p CPU transcode needs around 2,000 Passmark per stream. 4k needs "more", but I haven't seen data on how much "more". I haven't tried multiple 4k streams. When I stream one 4k movie to a phone, FreeNAS shows 95-97% CPU used by "nice", which'd mean there's still headroom. Plex doesn't show "throttled" transcoding, which means it's not significantly ahead, unlike 1080p transcode. That's with an e3-1230 v6, so pure CPU transcode and a Passmark of just under 10,000. It'd make sense that a 4k stream needs roughly 4 times the CPU power of a 1080p stream, but I don't know that's so.
Edit: Seems that's so. I can't keep two transcoded 4k streams going with that CPU. One is fine; two is a no go.

Then there's the HDR to SDR tone-mapping during transcode, and I don't know which software, if any, does that well. It may make sense to just keep a 1080p copy around for streaming to phones and tablets where a transcode is needed.

I didn't know to look for hw transcode when I built this, that's why I keep showing up in threads to make people aware of it. Whether it's a build consideration depends entirely on personal use preferences.

Streaming audio files is a different beast. Those don't need to be transcoded. You'll have your mp3 and that gets served to the mobile device which does the decoding. Not sure how other formats such as Ogg are handled, it'll depend on what the mobile device can do.

 

[LimeCrusher]

Is an RPi powerful enough to transcode 4K on the fly?

Not even close. Though the Raspberry Pi can be used as a Plex client with extensive audio/video codecs support hence nullifying the need for transcoding on the server side.

I've never seen The Wire

Literally stop what you're doing right now and watch it please.

 

[Platy]

Hi,

TL;DR I chose the Noctua AF14-PWM fans.

The difference between the AF15 and AF14 fans is that the AF15 uses 120mm fan mounts, the NF-AF14 should also have slightly higher max rpm. The Phanteks were not available at the usual online stores, the corsair 140mm maglev was more "power" than I needed and too loud in comparison to the others, the Bequiet silentwings are the most silent, but I feared the static pressure and air flow was not enought so in the end I went for the "most balanced" Noctua AF14-PWM fans.

[1]http://www.silentpcreview.com/article1345-page7.html
[2]http://www.silentpcreview.com/140mm_Fan_Roundup2
[3]https://www.quietpc.com/140mmfans

A few remarks:

-The included noctua anti vibration mounts and also the included screws are not compatible to mount them as "HD fans" with the fractal R5 Design case. Fortunately I replaced the existing fan and reused the screws and another set of 4 screws for the second HD fan were included in the "fractal case accessory package". The included screws however work fine as rear cooler.

-I connected them in the following way:

FAN1 -> back exhaust (low speed adapter)
FAN4 -> CPU (low speed adapter)
FANA -> HD FANs (low speed adapter+ splitter)

I set the correct fan thresholds with the ipmitool and changed the fan mode to "optimal" which should be the optimal mode for this setup. I chose to connect both HD fans to the same FAN "port", because in my opinion it does not make sense to have different fan speeds.

The (redacted) result is:

Code:

How many whole minutes do you want between spin checks?

NOTE ABOUT DUTY CYCLE (Fan%0 and Fan%1):
Some boards apparently report incorrect duty cycle, and can
report duty cycle for zone 1 when that zone does not exist.

Key to drive status symbols:  * spinning;  _ standby;  ? unknown                              Version 2018-01-01

Monday, Feb 11
          ada1 ada2 Tmax Tmean  ERRc CPU  FAN1  FAN2  FAN3  FAN4  FANA Fan%0 Fan%1 MODE 
22:19:06  *27  *27  ^27  27.00 -6.57  35   300   ---   ---   300   300    20    20 Optimal
23:00:36  *31  *32  ^32  31.50 -2.07  61   800   ---   ---   600   300    44    20 Optimal
23:41:07  *31  *32  ^32  31.50 -2.07  60   800   ---   ---   600   300    44    20 Optimal

The max cpu temp was 60-61 and the max HD temp 32 (cpu and HD stress testing), looking at the fan speed in my opinion a great result. The fans are so silent that I had to move my notebook because he was too noisy. Without the stress testing CPU temp is 40-46 and the hardisks 30 celsius, with all fans at 300rpm.

ISSUE: At the beginning I had some very strange issues, FAN1 spinning up to max rpm or showing as rpm speed 28800 etc also the various fanscript(glorius1,stux..) seem to require some change if you use a X11 instead of a X10 motherboard need to investigate further. I fixed the issues by resetting the bmc a few times changing the fanmode and change it back to "optimal" and finally the rpm numbers were "stable".

Please let me know if you are interested that I keep this thread updated was also thinking to create maybe a new thread? Not sure how appreciated it is to post blog style posts in this subforum. As I'm a freenas and NAS in general noob I'm going to try out a lot of different things and probably encounter a lot of problems (e.g. if you want to run ubuntu as linux vm there is a tricky "bug" regarding the uefi bios), finding the best cloud provider for my backups, best backup solution for all the devices (apple, google, ...) rsync? some opensource solution? backup to external harddisk.... I'm also in the process to upgrade my network: apu2 router with opnsense and ubiquiti hd nano and 8 port ubiquiti switch are going to replace my fritzbox 7490 box, I want to try out (in the near future) LACP, VLAN's, 10GE, WAN failover and so on.

sorry for the long text and my poor english :D