February 2, 20242 yr Hi all! So I want to use GPU transcoding to ram to avoid unnecessary SSD wear. So I follwed spaceinvader's tutorial for nvidia gpu transcoding and this one for ram transcoding This is my plex docker: Extra Parameters: --runtime=nvidia --mount type=tmpfs,destination=/tmp,tmpfs-size=20000000000 --no-healthcheck And in Plex: GPU transcoding works fine but I can't see where it's going to. I went to /tmp and can't see a transcode folder? Any help would be appreciated! Thank you!
February 2, 20242 yr Community Expert Cannot see any entry in the docker settings that maps the /tmp location that you have set at the Plex level to a location on the Unraid host (e.g. mapped to something like /tmp/plex) so the transcode temporary file will be inside the container.
February 2, 20242 yr Author Hm. I thought there was no need for that but that makes total sense. I need to add a new path then right? How would I go about it? Thank you btw!
February 2, 20242 yr Author 24 minutes ago, itimpi said: Cannot see any entry in the docker settings that maps the /tmp location that you have set at the Plex level to a location on the Unraid host (e.g. mapped to something like /tmp/plex) so the transcode temporary file will be inside the container. Actually I think I figured it out. I added a new path. Container Path: /transcode , Host Path: /tmp and it shows this: looks to be correct right?
February 2, 20242 yr Community Expert 2 minutes ago, Sptz87 said: Actually I think I figured it out. I added a new path. Container Path: /transcode , Host Path: /tmp and it shows this: looks to be correct right? I would suggest mapping it to a folder with /tmp as otherwise Plex may end up removing files under the /tmp path at the Unraid level that belong to other applications.
February 2, 20242 yr Author 3 minutes ago, itimpi said: I would suggest mapping it to a folder with /tmp as otherwise Plex may end up removing files under the /tmp path at the Unraid level that belong to other applications. I'm sorry, I'm not sure I follow. I'm absolutely new at this so I apologize for the annoyance! As you can see in the image that is within the container, outside the container /tmp does not have a Transcode folder. I thought this was all happening exclusively within the container.
February 2, 20242 yr Community Expert Simply mapping transcoding to /tmp at the Plex level inside the container does NOT map it to RAM, but to a location inside the docker.img file. What you would normally want is something like /transcode at the Plex level mapped to /tmp/transcode at the Unraid level since at the Unraid level /tmp IS in RAM.
February 2, 20242 yr Author 5 minutes ago, itimpi said: Simply mapping transcoding to /tmp at the Plex level inside the container does NOT map it to RAM, but to a location inside the docker.img file. What you would normally want is something like /transcode at the Plex level mapped to /tmp/transcode at the Unraid level since at the Unraid level /tmp IS in RAM. I think I understood? When I read your previous post I immediately changed the path config to this. As well as in temp directory in plex transcode settings to /tmp/plex_transcode, just to isolate it from the rest of /tmp Thing is, it shows it's mounted on tmpfs inside the container, as you've said. I can also see /tmp/plex_transcode at unraid level but nothing inside it when transcoding, only inside the container whilst I'm transcoding. Shouldn't this be theoretically correct and safe (as it's nested in its own folder) unraid level: root@xx:/mnt# df -h /tmp/plex_transcode/ Filesystem Size Used Avail Use% Mounted on rootfs 16G 1.2G 15G 8% / inside plex container: root@xx:/# df -h /tmp Filesystem Size Used Avail Use% Mounted on tmpfs 9.4G 172M 9.2G 2% /tmp
February 2, 20242 yr Community Expert Solution With those settings at the container level, you want to tell Plex to transcode to /transcode. The docker system will then redirect that to /tmp/plex_transcode at the Unraid level.
February 2, 20242 yr Author 29 minutes ago, itimpi said: With those settings at the container level, you want to tell Plex to transcode to /transcode. The docker system will then redirect that to /tmp/plex_transcode at the Unraid level. Perfect! Thank you so much!! That works! With the previous setup of just "/tmp" I had, I transcoded a couple of times to test, are the odds high that Plex screwed up something inside /tmp during that time?
February 2, 20242 yr Community Expert 27 minutes ago, Sptz87 said: With the previous setup of just "/tmp" I had, I transcoded a couple of times to test, are the odds high that Plex screwed up something inside /tmp during that time? That location is only inside the container so you should be fine.
February 6, 20242 yr Author On 2/2/2024 at 7:13 PM, itimpi said: That location is only inside the container so you should be fine. I'm sorry for reviving this thread but just wanted to ask another question related to this if that's ok! So I'm testing this out, have plex on my android phone, connected to the same network. I'm playing a large 4K 60 Mbps file, direct play to the phone. Absolutely 0 problems, super smooth, fast seek times, etc When I change the quality on the phone itself to say, 1080p 8Mbps. I get much more buffering and slower seek times than with direct play. On the same network. I honestly assumed this shouldn't be the case as the GPU would be fast enough as well as the bandwidth requirements would be way smaller? At the same time, it looks like it's barely affecting the GPU. Just hovering between 0-8% on Encoder / Decoder usage. Is this expected? It also seems to not change the size of the /tmp/plex_transcode/ almost at all, though I do see all the transcoded sections being created there. Just want to gauge if this is the expected performance on a setup like this? Specifically with such a powerful machine I was hoping it would eliminate most of these hurdles! +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 550.40.07 Driver Version: 550.40.07 CUDA Version: 12.4 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce GTX 1660 Ti On | 00000000:07:00.0 Off | N/A | | 44% 36C P2 24W / 120W | 451MiB / 6144MiB | 3% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 26040 C ...lib/plexmediaserver/Plex Transcoder 447MiB | +-----------------------------------------------------------------------------------------+
December 13, 20241 yr On 2/6/2024 at 6:48 PM, Sptz87 said: I'm sorry for reviving this thread but just wanted to ask another question related to this if that's ok! So I'm testing this out, have plex on my android phone, connected to the same network. I'm playing a large 4K 60 Mbps file, direct play to the phone. Absolutely 0 problems, super smooth, fast seek times, etc When I change the quality on the phone itself to say, 1080p 8Mbps. I get much more buffering and slower seek times than with direct play. On the same network. I honestly assumed this shouldn't be the case as the GPU would be fast enough as well as the bandwidth requirements would be way smaller? At the same time, it looks like it's barely affecting the GPU. Just hovering between 0-8% on Encoder / Decoder usage. Is this expected? It also seems to not change the size of the /tmp/plex_transcode/ almost at all, though I do see all the transcoded sections being created there. Just want to gauge if this is the expected performance on a setup like this? Specifically with such a powerful machine I was hoping it would eliminate most of these hurdles! +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 550.40.07 Driver Version: 550.40.07 CUDA Version: 12.4 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |=========================================+========================+======================| | 0 NVIDIA GeForce GTX 1660 Ti On | 00000000:07:00.0 Off | N/A | | 44% 36C P2 24W / 120W | 451MiB / 6144MiB | 3% Default | | | | N/A | +-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=========================================================================================| | 0 N/A N/A 26040 C ...lib/plexmediaserver/Plex Transcoder 447MiB | +-----------------------------------------------------------------------------------------+ Something is definitely not right. I'm using a GTX 1660 SUPER for all my GPU assisted work in my server, capabilities should be the same as your 1660 Ti? For reference, I'm transcoding 2 video files with Tdarr and transcoding a 4K DV/HDR file to my phone over 5G Same Plex file over WIFI and 1 transcode with Tdarr (The other one had finished in the meantime) And just for shits and giggles, only Plex transcode, same 4K DV/HDR movie over 5G I noticed your encoder/decoder is hardly working. Are you sure you have the right read/write access to your temp/transcode folder?
December 13, 20241 yr Community Expert The GPU throttles after the transcode buffer is filled. I can't remember what the default setting is but it's either 60 or 120 seconds. Once that buffer is met, the transcoder throttles and doesn't need to encode/decode but in small segments. When you start seeking, the buffer is nuked and has to rebuild. Based on the screenshots, everything looks normal. At the very start of a transcode session you will see the Encoder/Decoder stats climb up but after the buffer is full, it will drop to 1% or 0% until more buffer is needed. Pay attention to the throttle state: P0 is most power <---> P8 is idle. During Transcode Buffer Filling: Buffer is full: Edited December 13, 20241 yr by MowMdown
January 19, 20251 yr On 12/13/2024 at 12:12 AM, Degn said: Something is definitely not right. I'm using a GTX 1660 SUPER for all my GPU assisted work in my server, capabilities should be the same as your 1660 Ti? For reference, I'm transcoding 2 video files with Tdarr and transcoding a 4K DV/HDR file to my phone over 5G Same Plex file over WIFI and 1 transcode with Tdarr (The other one had finished in the meantime) And just for shits and giggles, only Plex transcode, same 4K DV/HDR movie over 5G I noticed your encoder/decoder is hardly working. Are you sure you have the right read/write access to your temp/transcode folder? Sorry for the minor thread necro, but don't these screenshots show you using your CPU's transcoder? (Epyc) and not your GTX renderer? sorry- I came back to this thread after updating to 7.0 and realizing id long forgotten to pass /transcode to /tmp/plex and when reading the thread I noticed this and wondered
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.