ich777

Das passt perfect sobald die CPU QuickSync unterstützt kannst du in Emby, Jellyfin, Plex transcoding auf die GPU auslagern. Welchen Container bentutzt du, sprich von wem, Repository? Wenn du zB auf einem Mobilgerät Medien ansiehst die dein Mobilgerät vom Format nicht unterstützen muss das Medium konvertiert werden und das geschieht normalerweise in Software sprich auf der CPU und da die CPU nicht für Videokonvertierung optimiert ist kannst du dir vorstellen das du viel Last verursachst. Wenn du aber die integrierte iGPU nutzt mit QuickSync nimst du die Last von der CPU weg und die integrierte Grafikeinheit übernimt das konvertieren, viel effizienter, schneller und natürlich auch Energiesparender. Dein Prozessor sollte dafür mehr als ausreichen. Du verwechselst hier was, es wird immer am Server transcodiert und nicht auf den Clients, wenn du auf dem Client eine andere bitrate auswählst wird am Server konvertiert und dieses konvertierte Format am Client wiedergegeben. Wenn du es versuchen willst lad dir am besten mal mein Intel-GPU-TOP Plugin (nicht den Docker container) runter und das GPU Statistics Plugin von @b3rs3rk damit du die Auslastung deiner iGPU siehst (beim GPU Statistics Plugin musst du nochmal in die Einstellungen rein und deine Intel iGPU auswählen). Danach gehst du in dein Emby template und fügst das hinzu: (danach sollte der Docker wie gewohnt starten, wenn nicht wurde deine iGPU nicht richtig aktiviert lösch dann einfach die oben angelegte Device wieder - nur als Hinweis) In Emby gehst du dann ins Server Dashboard und bei "Transcodiere" stellst folgendes ein: ...und aktiviere auch den Punkt: Wenn du danach ein Video startest und am Client die Qualität runterstellst (kleines Zahnrad -> Qulität und auf 480p oder so stellen, sollte hald niedriger als das Original sein) kannst du im Server Dashboard sehen das in HW transcodiert wird (VAAPI) und du fast keine CPU last mehr hast (fast deswegen weil Audio, Untertitel usw auch noch konvertiert werden müssen und das geschieht auf der CPU). Hoffe das beantwortet deine Frage(n).

This is common and depends on how the manufacturer implemented the default fan speed, if you are only transcoding with this card the fan speed wont change that much since transcoding produces only a fraction of the load that the card is capable of and the fan simply won't ramp up because the card doesn't get hot enough... Yes on some cards you actually can't see the power usage that is reported from the card and it can also happen that on a newer driver that it doesn't work anymore because Nvidia breaks often the compatibility to other or should I say older cards. For example my GTX 1060 3GB reports everything, then my 1050Ti isn't reporting the power usage as in your case, but there is actually nothing I or @b3rs3rk can do about that. Hope that makes sense to you.

Das wäre unter Umständen mit dem GVT-g Plugin möglich, jedoch musst du bedenken das du je nach Hardwarekombination ein Limit an gleichzeitig laufenden vGPUs vorliegt. Wenn du eine Display Ausgabe willst ist das bedingt mit dem GVT-g Plugin und vGPUs auch möglich jedoch musst du dir zB einen USB DisplayLink Adapter besorgen, diesen an die VM weiterreichen, Treiber installieren und dann die Ausgabe von deiner vGPU auf den DisplayLink adapter spiegeln, somit hättest du auch ein Bild (natürlich mit HW Beschleunigung) mit nur einer richtigen GPU (Intel iGPU). Jedoch kann ich dir nicht sagen ob es reicht nur den DisplayLink Adapter durchzureichen oder ob du einen ganzen USB controller durchreichen musst an dem der DisplayLink Adapter hängt.

Really strange but glad that it is working now...

Danke... Das kann durchaus korrekt sein wenn das Modul nicht größer ist... Ich hab jetzt aber nur ganz überstürzt heute Früh das Repo nochmal durchsucht und bin auf folgendes gestoßen: Klick und natürlich hab ich das ganze ohne die Firmware Dateien kompiliert, die anscheinend nötig sind aber wie gesagt ich hatte noch nicht wirklich Zeit mir den ganzen Issue durchzulesen ob das überhaupt noch notwendig ist aber ich schätze schon da es im Readme vom default Branch drin steht: Klick.

Exactly. This won't work by now, please don't try it any further, this is a dead end for now. I will report back if I got some news for you.

Is really difficult on Unraid but I will investigate further. A dummy plug and a DisplayLink device are two different things. A display link device is like a USB graphics card with a HDMI or VGA or whatever output on it and you will only need it if you want to output the video signal. You simply have to pass the DisplayLink device through to the VM and mirror the GVT-g iGPU screen so that you get a output hope that makes sense... A dummy plug for the iGPU is only needed if your iGPU isn't initialized properly when booting Unraid.

You actually can test your server first here: https://webrtc.github.io/samples/src/content/peerconnection/trickle-ice/ Can you reach it from there? What ports have you forwarded? Have you changed anything in the template?

Can you try it once more without persistence mode? You have to reboot to actually unload the persistence mode or kill the process 'kill "$(pidof nvidia-persistenced)"' (without the single quotes from a terminal), but I think the better choice would be to reboot. I don't think that an old plugin caused this, eventually you need persistenced mode enabled.

Do you mean for Unraid as the main graphics output (console or Unraid GUI mode)? Then yes. Or do you mean to display the output from the VM? Then no, at least partially... You can set up a DisplayLink device (for exemple a USB one), install the drivers in the VM and set the Display to "mirror output" then you have the abbility to display the output on a physical monitor with the accelerated vGPU. Hope that makes sense to you.

I have to investigate a little more, since this is really strange. I had a user here I think it was @Glasti that had a 1660 and a B450-f but it was a different error if I remember correctly. It seems really strange since from what I see in your syslog it at least tries to transcodes and then fails I think, not 100% sure why there are so many BAR messages... What you can also try is to turn on persistence mode to do that open up a terminal and enter: nvidia-persistenced ...and then try to transcode some of your media (eventually restart the container after you enabled persistence mode), if that will also fail I really can't help here and only can think of something like a hardware related issue.

Eventually but I'm not really sure since I'm not a OSX user. @giganode can you help out or at least share you opinion if that's possible? The best place to ask would be on the macinabox support thread. I don't think it would be any different from the tutorial above until to Step 4, Step 5 would be to get the vGPU working in OSX.

Please force an update of the container and try to remove the TURBOVNC_PARAMS variable again, have changed the Chrome container and now it should have no such collisions. The other containers with VNC are currently rebuilding now (Dockerhub is really slow with the uploads today about 800KB/s... )

But is a monitor connected? Is the card also passed through to a VM?

The log seems fine and it seems like it's loaded the driver correctly. I think you boot in Legacy/CSM mode from what I see on your boot device which seems fine. Please check if there is an update of the BIOS and keep me updated will research a little more. Eventually try to connect a monitor if you haven't one connected.

May I ask if you run a custom Kernel or something else? Can you make a reboot and then post the diagnostics again. I also had other users with a AMD/B450 boards where the cards aren't working, can you eventually check if there is a BIOS update available and update the BIOS if possible?

Have you restarted your server/Docker service after installing the drivers? In Jellyfin you can tick all boxes (MPEG2, MPEG4,...). Something seems really of if you get the error. Have you already tried the Production Branch drivers (anyways "latest" also works fine because I run that on my system)? Is the power connected correctly to your GPU if there is any additional power plug? Can you send me the Diagnostics before changing anything (Tools -> Diagnostics -> Download -> drop the downloaded zip file here in the text box).

The 750 is a supported card and the plugin should work with it just fine. From the two comments that you've made you actually don't need the Nvidia-Driver plugin since the plugin is only a addon for Docker containers to use the cards in containers and as you said you use them in VM's you have not much of a benefit installing it. You can of course use the 10900K with QuickSync (I'm also using it on my 10600 and from a technical point the QuickSync engine is on both CPUs the same). It should be capable of transcoding 4 to 5 simultaneous HEVC streams (depending on the settings) without a problem, also like the 1660. That depends on what you prefer but you can't use the 1660 in a VM and also for a Docker container at the same time. You setup would be perfect if you do it like that: 750 for VM 1660 for VM iGPU for Docker If you do it like above you only have to install the Intel-GPU-TOP plugin (the plugin handles the installation and enabling of the iGPU and also sets the permissions), in Plex add the device: '/dev/dri' (without qutoes) and you should be good to go (and of course you can use the output of the Intel iGPU for your Unraid terminal/GUI - please also make sure that you actually set the primary video card to the iGPU in the BIOS and connect a monitor or dummy plug so that the iGPU is properly enabled when you boot up the machine). What I also recommend is to uninstall the Nvidia-Driver plugin since you don't have much benefit if you are using the Nvidia cards in VMs and not in Docker containers. Hope that helps.

Please install my Jellyfin container as said in the above post (since I know that it is working). Have you set all the necessary variables? Please also send a screenshot from the template.

Can you eventually try to install my Jellyfin container and add the necessary variables and try to transcode a file? Only for troubleshooting reasons... Description is actually in the second post.

What card do you own and what type of content did you try to transcode?

What container are you using, better speaking from which repository/maintainer? Have you read the second post of this thread and added all necessary variables and parameters?

This is not possible from what I know. You need nvidia-settings for that and nvidia-settings depends on a GUI or better speaking X server. You can set power limtis for the card with nvidia-smi but not fan speeds.

This is not my plugin My plugins are the Intel-GPU-TOP, Radeon-TOP and Nvidia-Driver... About which iGPU are we talking? Both Intel (8th Gen.+) and AMD (Ryzen APUs) have excellent iGPUs that should be capable of transcoding the same amount of parallel streams as the 1660 with the benefit of not having a base power consumption when they are in idle like the Nvidia GPU has. Thank @b3rs3rk, the creator of this wonderful plugin.

What do you mean exactly? I have had a little more specific there, it was asked a lot of times to display multiple cards in the plugin and most people thought that the plugin can also display the utilization of card that are handed over to VMs too but that is simply not possible. What's your use case that you have multiple GPUs in your Unraid server installed? For VMs I think or am I wong? I don't think that many people have two GPUs in the system that are both handed over to Docker containers.