Emby Theater PEGS My GPU on MY Intel NUC

[scubajwd 0]

As the title suggests running Emby Theater ON the Emby server machine (Intel NUC6i7KYK..aka Skull Canyon)

eventually pegs my GPU causing the NUC to shut itself down from overheating.  It is a very

repeatable issue with my rig,BTW..does this for both the released version of Emby server and the latest beta

release as well..before I posted this I;

 

1)updated all the Intel NUC drivers and bios

2)toggled off hardware acceleration in the Emby server

3)removed a plethora of other client server apps to try and isolate

the problem: Plex, dropbox, and Teamviewer to mention just the biggies

4)removed completely the Emby beta install and installed from scratch the

released versions of Emby server and Emby Theater

5)I monitor system load and temps with HWMonitor app and task manager

6) Running Emby Theater on the LAN DOES NOT produce the issue; just when I

run it on the main Emby server

7)During the GPU pegging (goes to 100%) the cpu remains relatively low < 15% but

the system becomes unresponsive and has to be manually hard stopped..system auto

shuts down when a thermal limit is hit..NUC case is extremely hot at this point..

 

I've run Emby theater on the NUC in the past w/o this issue...so either an Intel NUC upgrade

has caused this and/or a Win10 upgrade; I've checked the fan inside and blew out any dust and I try to add

cooling to the room this little beauty lives in...

 

Currently recording two shows and direct playing one to my Roku ULtra in my living room w/o

issue 

 

I could have a faulty GPU that is on the verge of failing..don't know ..running Emby server stress tests

as I write this..so I'll be back if I learn anything else

 

Anyone have any ideas or things to try ; this NUC has been rock solid for me with Emby for a long time..

[Luke 37252]

Hi there, let's look at an example. Please attach the information requested in how to report a media playback issue. thanks.

[scubajwd 0]

Luke, the runnup in GPU happens so fast that I cannot get a screen shot of the HWMonitor

output; my last test failed so my problem is NOT just related to Emby Theater..

By failed I mean the system auto ejected at a thermal limit..right now I'm just

running Emby server w/o any recordings or playbacks going on..stay tuned..Emby has

been rock solid on this rig for a long time..something within the app is triggering

a hardware event that I've not seen before..

[Luke 37252]

The server logs will help tell us how Theater is playing. That's the first thing we want to identify.

[scubajwd 0]

Well..it appears that I had a couple of problems with my rig;  I took the NUC apart and

again blew pressurized air thru out the fan assembly and particularly the outboard

cooling fins; this made a huge difference dropping the idling fan speed down to 1000 rpm

from 1500 rpm w/o load; I then (on a hunch) eliminated an iscsi connection to one of my NAS units that I

use for transcoding instead using a faster external ssd; this eliminated a playback bottleneck as

the cpu % climbed as the slower iscsi connection could not keep pace; all of this has eliminated the

GPU/temp spikes that were crashing my system; it seems,however, that the Emby Theater app

exerts a load on my server that I never understood before and makes it run hot; I'll rerun some

of those tests now with the iscsi connection eliminated..that will help I'm sure..  I'm still

wondering why the configuration worked w/o issue before..I didn't have the hw monitor installed

and running so maybe never realized how close I was to the NUC thermal thresholds

[Luke 37252]

Thanks for the feedback.

[scubajwd 0]

Luke, will your upcoming 3.6 release use less server

resources than 3.5?  Will the under-the-hood updates

be more server friendly?  Both Emby & Plex seem to

demand a pretty beefy server for varied household use of

recording & playback activities..

[Luke 37252]

Yes that is correct. The server will use less resources.

[scubajwd 0]

On chat support call for about an hr today with an Intel Tech; we ran some stress tests on my hardware

and could not get it to crash from reaching a thermal threshhold; 105 Deg C (believe it or not) is the

tJunction threshold for thermal induced shutdown  (>212 Deg F BTW)..so its hot..so I've stripped off all iscsi connections

and external drives and will restart Emby tests; I will likely try the latest  Emby  beta for this; will keep you posted

[Luke 37252]

Ok, I've mentioned log file examples a couple times already. It will be difficult for me to provide very much insight without that. Thanks.

[scubajwd 0]

Been running more tests; don't want to burden you with MY problems; will be testing

latest beta today on my simplified NUC( both server duties and Theater) will provide

logs then..early testing,however, with latest beta is quite positive..will

run Theater for 45 minutes on  main Emby server to see if it causes overheating

via a runup in GPU..not so far,BTW

[Luke 37252]

Thanks for the info.

[scubajwd 0]

Running very smoothly on beta 4.0; combination of my hardware mods and version 4.0 seems

very stable now on my NUC; temps and fan speeds way down from previous test runs; well below

temp thresholds now and no sign of GPU heading towards infinity..machine resources with 4.0 seem

way less than previous beta versions but I have changed my hardware a bit so it might be hard

to make comparisons..

[Luke 37252]

Thanks for the feedback !