I have an ODROID-XU3 ARM device that I am running ArchLinux on and the Plex Media Server.
As you may know the XU3 is an 8-core samsung device and I have found that it is totally sufficient to transcode even the worse 1080p divx or other content to the web client.
(Other clients like Android or the PHT that do transcoding on the client hardware have been eliminated from my testing)
So I and others are annoyed at the "This server is not powerful enough .." message that prevents transcoding the above content.
(In testing I have disabled that test by inserting a "return false" in the validateTranscoder() function in the plex.js script)
I have tested as much content as possible to know that this hardware is capable of the most demanding transcode demands. The 8-cores are hardly used for most content and at it's worse
the CPU's are at 70% on 4-cores, the other ones <10% .. (Another user I have been in contact with has the same results as me ..)
So can I please request that you re-look at how this function determines the hardware is good enough.
At the moment every update over-writes the changes I do to this file, so I have to edit it frequently to use my system again to it's full potential..
I have put this in the “Feature Request” section, but no developer has had the decency to reply, … yet …
It’s nearly 2 months since posting … and silence …
Would be great if Plex would run on our shiny odroids without any problems!
The plex.js “hack” will not work for other clients that the web interface? I am using my iPad when I’m on the road and would love to stream my media from home but it doesn’t transcode.
Any hints on how to get other clients working as well?
You might be able to use the api to turn on your transcoder (if there is a build for your cpu). Can you post the xml from http://server:port and http://server:port/:/prefs, make sure you blank out the machine id or any other identifying bits. I’ve used one of these settings listed to turn off my transcoding, but it didn’t work as expected (now that I think about it is was probably the faststart/web optimized header that was the issue), maybe we can try the reverse for you after comparing yours vs mine or anyone else’s. Though I would be surprised if you couldn’t doing the same thing through the config file. Post or pm that as well.
Passmark is the score used to measure the different CPU’s against each other. Such as how an i3 such and such against an i5 so and so. It’s used to determine the suitability of a given CPU for a known requirement or task.
In other words, its the score that tells the developers if making server side component of PMS for your particular device is worth their time or not. The CPU in your device isn’t even listed in most benchmark packages, including those that maintain the Passmark system: http://www.cpubenchmark.net
This CPU can generally handle 2 1080p 8Mbps rips to almost any down sampled bitrate. As the bitrate of the original goes up, the CPU passmark scores scale up as well. So if I tried to transcode a 20Mbps movie to say 4Mbps, the CPU is possibly going to struggle. It doesn’t now, but it could. Introduce bitmapped subtitles, and then you get into a realm of things that a lot of other machines can choke on.
In the realm of NASes, this is one of the more powerful for consumer grades, yet is a bare bones CPU for running PMS on a lot of other machines. I use it very successfully. Because I know how to make things work without a lot of excessive transcoding taking place. It’s not hard to do, but it is time consuming to set up or hand convert everything over. But anyone can figure it out or ask me or a slew of others what needs to be done to reduce transcodes.
This is where you are going to have problems convincing anyone that the small CPU in your device is going to be able to seriously handle what you are wanting to do with it. You may be able to use it for a player, and it might run PMS on it, but it’s doubtful that it can run even minimal transcoding sessions. You’ve already been asked for the passmark ratings on the CPU. Now you know how to get it, so we’re waiting to see the results.
This is where you are going to have problems convincing anyone that the small CPU in your device is going to be able to seriously handle what you are wanting to do with it. You may be able to use it for a player, and it might run PMS on it, but it’s doubtful that it can run even minimal transcoding sessions.
Here speaks someone who has not tried the Samsung BIG.little 8-cores for the task at hand.
Well I HAVE! … and it performs just great. It transcodes every format I throw at it to the web client, without so much as a pause in the stream … I have transcoded 1080p streams over a 1Mb slow interface just fine, or transcoded one 1080P format to another … no trouble. I have also served up 3 1080P streams with no issue …
Put your money where your mouth is and actually try it on the hardware in question … in my case an Odroid-XU3 … it’s one thing to mouth off about how it can’t do it, when you havn’t actually tried it, it’s another to KNOW for a FACT it can.
I asked about the Passmark because I only googled a commercial software by that name … and only for x86.
Your link did not show the download for any ARM version and especially one optimized for the BIG.little core as my server code is ( I compiled key software like PMS with optimum CFLAGS settings for the CPU) …
If passmark was opensource software, I would compile it for the Odroid and give you a FAIR result, right now you can’t even point me to an ARM executable of any sort, so how can I do what you ask???
And yet, this is the scale your machine is going to be judged by, not by what you say it can do, but by what those companies that benchmark the CPUs say it can do.
And I’m not interested in the least to “put my money where my mouth is” as you say. My hardware can do everything I want of it and a lot more. And the apps are already written for it. That’s the fight you face now…
Getting support for your device is going to require YOU proving the device is capable. You or anyone else interested in stepping up to help with the task you are taking on. That means you are going to have to PROVE the CPU can do the task by getting hard and fast numbers showing it’s abilities, and not “mouthing off” with unsubstantiated statements running on a hacked together application.
And that’s the fight you have in front of you. And in front of the little device you want to try running PMS and transcoding media on.
I KNOW for a FACT that I can run PMS on a device with under 300 passmarks. Because I HAVE done that. And I did that only by minimizing transcoding as much as possible. I KNOW for a FACT that 2000 passmarks is required as a bare minimum for transcoding an 8Mbps 1080p stream with most any main stream CPU’s out there. I KNOW this because I’ve first hand experience with it. What I don’t KNOW (and honestly don’t really care, one way or the other) is if YOUR CPU can do what the CPU in my NAS can do.
Now you have a CPU that when I look it up I can’t find any reliable benchmark data on. And almost every one of the reviews I find for it say it was used as last years low end cell phone or cheap tablet. That’s the info I can find on the CPU you say can transcode a 1080p stream with no problems.
So this comes down to one thing, then. Prove it otherwise. Prove, with hard and fast numbers from a reliable site that the CPU in your device is up to the task. This is on YOU, not on me.
I just finished telling you that there IS NO ARM passmark program, so how can I give you the passmark ratings … but you knew that didn’t you …all you want to do is start a war …
If you give me a benchmark program that I can compile/run on my ARM I will, especially if there are results that I can compare with x86. But you can’t … and that whole plex “minimum reguirments” link using passmark is only geared at x86 so the deck is stacked against me and others …
If you think your NAS can compete with the samsung core with 4-cores at 2Ghz and 4 at 1.5Ghz, good luck to you … like I said, I can transcode 3 x 1080p streams simultaneously to a lower bit-rate, can you?
I don’t actually have to prove anything. I am only asking developers to concider making the transcode a user switch, so that those like myself can transcode content quite happily and can do so for all clients, not just the web client.
Thinking that a “passmark” score is the bench to which all CPU’s must attain to be ALLOWED to transcode speaks to me of “control” and “manipulation”, rather than being open and transparent. Most users on Linux devices know enough about their system to make a wise choice about transcoding or not … Linux is all about “user” choice, plex has the danger of falling into the “microsoft” mindset of >we will decide what you like< … if this sort of attitude continues …
also … remember it’s not just about CPU horsepower … it’s a lot about how efficient the codecs work with the CPU architecure … i.e if the codeing/decoding can be efficiently shared across cores …
Well, no, I really didn’t know that… What I did know is that there needs to be some sort of benchmark to use as a guideline. It’s that way with any piece of software, even games… Imagine trying to play any modern game on an old Tandy 1000. It’s not going to work is it? Without a benchmark to tell someone the minimum specs of an app is, the more confusion people have when they try to use the app.
And yes, I do think my NAS can outperform your device. But you see, I don’t need to transcode three to a lower bit rate, because the lower bitrate versions are already made for 90% of my library, so they are almost all Direct Played at the lower bitrate as the client requested it. HDD’s are cheaper than a 20K passmark CPU is. So I chose to make multiple versions to support the client apps as needed. A single 3TB drive holds most of those other versions… That’s less than 10% of the whole pool of drives I have hanging on the NAS, so in my mind it’s a no brainer to have it on hand… In fact, almost none of my media is ever transcoded, except under special circumstances. And I usually correct that at the first opportunity to maximize my server’s operations.
But we have no way to actually test if your little device can outperform my NAS, do we? So all of the bluster and fussing gains you nothing but raising your blood pressure and makes others laugh even harder at the whole thing. I’ve already been getting a bunch of PM’s about this thread now, laughing at this and telling me to just let you be… It’s your sandbox and I should just let you play in it… You aren’t helping your situation at all, but I bet you think you are… And that’s the sad thing.
I don’t HAVE to give you an app to benchmark your device. You have to find one that is recognized in the industry, has quantitative results and can be duplicated across other similar devices. I actually hope that you get it someday… The more people using Plex in the future the more secure the suite is long term. What I care about more is if the developers start developing the app for you and drop working on the features that already have a large contingent of users voting for them. Those same folks that have been waiting for some of those features for a very long time…
At this point, I’m out of this thread. The minutes I’ve spent on this whole discussion are wasted as you won’t ever get any kind of support for your device and PMS going as you are now. It’s obvious that my suggestions to you aren’t going to be taken. So continue the blustering, and good luck with that feature request.
I don’t want them to support my ARM processor at all. All I want is the ability to make the transcoding decision on my device, not have it taken away just because they only want to support some select x86 devices …
Is that so much to ask? is that blustering?
Most of us “non” developers i.e users don’t have >30T drives with multiple versions of videos in different bit-rates. I am just wanting to use plex as I believe it was designed for, sending efficient streams over networks & clients of various qualities … plex is really good at that.
Anyway, what is wrong with running the server on an odroid-xu3/4 or a RPI 2 if it is quite happily transcoding, it would seem to me that it would be in plex’s best interest to be able to run the server on cheap hardware that runs at low power, the market is huge for example in $30 RPI’s … and not many people would choose a server using >100W 24/7 if a 4W odroid will do the job just fine … as mine has for the last 5 months … (and my power bill is much nicer thank you very much)
