This is a different issue. Please open a dedicated support thread with the details and appropriate logs from the system.
Yes. I realize it’s not the same issue. But I’ve had zero luck in getting support. I don’t know what logs they need and the zip file for all of the logs exceeds the 8MB limit.
As far as your issue, that’s experience I’ve had with all generations of QSV with the exception of my 11th generation NUC which ONLY records MPEG2 from cable. It looks fine converting MPEG2 to H.264. So it seems QSV is still a pile of junk.
Related to the quality limiting settings - I was helping a family member with the Tizen Plex app on their Samsung TV over the holidays. None of the content would play unless the Maximum Remote Quality setting was explicitly defined.
Leaving this setting to Unlimited / Maximum would result in a spinning loading icon and the following errors. The stream would never start. The content being tested had a maximum bitrate of ~7mbps.
Between the server setting for limiting the bitrate to increase the quality, and the client not playing content unless this is explicitly defined, I have no idea what’s going on anymore.
@ChuckPa, @chrisallen any ideas here? See my post 5 items up for the server settings. I don’t know how PMS determines ‘unlimited’ or ‘maximum’ for these settings. – I’m wondering if it has anything to do with the following: Available public upload bandwidth is 1gbps, The Intel NUC running PMS has a 2.5gb NIC, and the observed link between storage and switch is 10gb.
Can I have some Server Debug Logs which capture this happening, please ?
I have a few ideas but need to see what’s actually happening because we’re talking about a TV as the player.
The logs will show me:
- What’s actually in the video file
- What the host can do
- What the player (TV) can do
- What was decided when PMS & TV negotiated the playback settings
@chrisallen It’s been 6 months, can we please get an update on this? We just got the first GA PMS update in 2 months and the issue is still present.
I am getting ready to put together a Plex server right now. I will be running Ubuntu.
I would rather not have a discrete gpu. But after reading this thread and the other ones, I’m worried that the Intel igpu will cause problems.
Is there an Intel CPU that definitely does not have this issue but is plenty powerful?
Thanks in advance to anyone who responds.
I run a Xeon-D 1541. I can’t tonemap or transcode 4K, but anything else is fine for my personal usage. This issue is only for people doing hardware transcoding with iGPU.
1 Like
Is there an Intel CPU that definitely does not have this issue but is plenty powerful?
This thread discusses quality issues running PMS nested on a 11th - 13th gen NUC with intel Iris XE iGPU and on 12/13th gen CPUs with UHD 770 iGPU.
If you want to avoid similar problems, go 10th gen.
EDIT
The issues brought up in this thread have been tracked/updated in the first post. Avoid those settings and the transcode quality seems to be okay.
I suggest checking other threads for hardware recommendations specific to your use case and hardware. There are a ton out there.
1 Like
I’m still pending confirmation, but I believe the client issue to be resolved with the release of 1.40.0.7998
(Transcoder) HW encoding would fail on devices with no rate control (#14222)
I can confirm there are still playback quality issues when setting “Limit remote stream bitrate” to "Original (No Limit) in the server settings. (See my Update from October 2023)
2 Likes
Thank you! I will do that. Glad I read through here before spending any money.
That xeon doesn’t have QSV
ALL:
I’ve been doing a lot of reading about the Intel 770 and XE machines.
I’ve also come to understand how some of you are building/configuring them.
Some of the smaller CPUs with XE capabilities in them (like the N100) come with severe limitations & cautions.
Findings:
-
XE graphics really does not work well unless Dual Channel mode (128 bit) is active
– Two identical-spec memory DIMMs installed
–sudo dmidecode -t 17confirms dual-channel is active -
“cheap” XE CPUs (like the N100) don’t have the clock speed/bus speed to to drive for higher bitrate like their desktop counterparts. ( a Beelink Jasperlake is faster )
– Transcode decode
– Tonemap
– Transcode encode
I’m still working on this but I think I’m seeing a trend where folks are taking one path because of a YouTuber or something – and that info is incorrect / incomplete for all use cases
Thank you for your investigation - I don’t know anything about XE being limited by the memory installed - I don’t have to worry about that since I have 64gb dual channel with the Iris XE EU. (I believe it’s branded as Iris XE MAX now)
Not sure what you mean by folks taking a path because of a youtuber though, or what info might be incorrect / incomplete. A NUC is the perfect platform for labs and containers, and running plex in a container on it with pass-thru just makes sense. - Why waste all that compute / memory on just plex when I have 10 other things that need to run next to it. I’ve dedicated 32gb to plex and transcode straight to memory which leaves me another 32g free for other things. With the container it’s running on now, I can get around seven local 1080p transcodes before buffering starts. There’s a bottleneck somewhere, but I just don’t need that many streams and haven’t investigated it further. This issue wasn’t just related to NUCs though. I did reproduce it on a bare-metal 13600K. But I digress-
In relation to the original problem, the quality has been fine for a few months now. I just need to keep the LAN box empty and keep all streams rate limited instead of leaving it uncapped. (server settings) - Whether that’s acceptable to others- I’m not sure, but it still seems like an issue that needs to be investigated.
For purposes of sharing with others who might not know, would you mind showing the sudo dmidecode -t 17 output so they can see and understand how the dual channel is listed?
I’m finding branding doesn’t mean anything. CPU specs say a CPU is capable but unless memory is correctly configured , the “branding” isn’t “available” to use.
Here is the output
Thanks.
Here’s the other way it works:
[chuck@lizum ~.2001]$ sudo dmidecode -t 17
# dmidecode 3.3
Getting SMBIOS data from sysfs.
SMBIOS 3.1.1 present.
Handle 0x0030, DMI type 17, 40 bytes
Memory Device
Array Handle: 0x002F
Error Information Handle: Not Provided
Total Width: 64 bits
Data Width: 64 bits
Size: 32 GB
Form Factor: SODIMM
Set: None
Locator: ChannelA-DIMM0
Bank Locator: BANK 0
Type: DDR4
Type Detail: Synchronous Unbuffered (Unregistered)
Speed: 2667 MT/s
Manufacturer: Samsung
Serial Number: 0161034D
Asset Tag: 9876543210
Part Number: M471A4G43MB1-CTD
Rank: 2
Configured Memory Speed: 2667 MT/s
Minimum Voltage: 1.2 V
Maximum Voltage: 1.2 V
Configured Voltage: 1.2 V
Handle 0x0031, DMI type 17, 40 bytes
Memory Device
Array Handle: 0x002F
Error Information Handle: Not Provided
Total Width: 64 bits
Data Width: 64 bits
Size: 32 GB
Form Factor: SODIMM
Set: None
Locator: ChannelB-DIMM0
Bank Locator: BANK 2
Type: DDR4
Type Detail: Synchronous Unbuffered (Unregistered)
Speed: 2667 MT/s
Manufacturer: Samsung
Serial Number: 01610349
Asset Tag: 9876543210
Part Number: M471A4G43MB1-CTD
Rank: 2
Configured Memory Speed: 2667 MT/s
Minimum Voltage: 1.2 V
Maximum Voltage: 1.2 V
Configured Voltage: 1.2 V
[chuck@lizum ~.2002]$
The goal is to not transcode 4K to start with.
Here’s a screenshot of me transcoding 17 Mbps 1080p HEVC, with lossless stereo audio and subtitles, with no hardware acceleration, and using the highest transcoder quality setting:
Yeah I agree. I had to force transcoding for that test.
I am still interested in this topic, though. Because I am thinking of separating my Plex server from the rest of my NAS and moving it to an N100 mini PC, so I’ll be using a 12th gen Intel and QuickSync then.
From first-hand testing –
Don’t use the N100. It won’t cut it. I helped a user test it. He thought it would work too. It fell FAR short. (Internal clock / bus speed can’t keep up with demands of pushing around raw 4K video frames for tone mapping) … It stuttered badly eventhough the HW was working
Go get the JasperLake
chuck@jasper:~$ grep 'model name' /proc/cpuinfo | uniq
model name : Intel(R) Celeron(R) N5105 @ 2.00GHz
chuck@jasper:~$
What again are the conditions that will cause an N100 to fall far short? I have not ran into that issue at all yet and would like to do some more testing. Is it only when tone mapping? I have not really tested that yet as all of my clients either don’t need it (AppleTV with HDR sets or Plex) or do it themselves (AppleTV on non-hdr set).
