First off I love Plex and have been a power user for many years now. It’s a bit disappointing though to see that in a few key performance areas Plex is getting beat out by other solutions. Emby currently has 4 different types of Hardware Transcoding supported, Intel Quicksync, OpenMAX OMX, Nvidia NVENC, and VAAPI.
With newer intel CPUs supporting Quicksync and VAAPI broadly (even lower power devices with Atom CPUs) I think this is a huge opportunity for Plex. How many people have seen the threads on here about people wanting lower power Plex servers and asking about what the best build would be? These different types of hardware transcoding anecdotally make a huge difference in performance, with some people being able to cut CPU usage down by a factor of 5 during transcoding.
Not only would this be huge for lower power devices, this would be huge for people managing big Plex libraries and setups (like myself).
It would also lower the bar for a lot of people to jump into running a Plex server, which would be better for the company as a whole.
This IS kind of a feature request, but I would love to have a larger discussion about this with all of you instead of just saying “I want this, please make it happen”.
The Nvidia Shield (presumably NVENC) and WD My Cloud PR2100/4100 (presumably Intel QuickSync) both already have hardware encode acceleration. I expect it’ll be coming to other platforms sooner or later (alas, my money’s on later).
Great topic and good question. Secretly I hope that the Shield/NVENC and the WD-machines (Quicksync) are trials/test-cases with the goal of supporting more machines/chips with the specific featuresets.
@mickeldaelmans said:
Great topic and good question. Secretly I hope that the Shield/NVENC and the WD-machines (Quicksync) are trials/test-cases with the goal of supporting more machines/chips with the specific featuresets.
Wait… So plex is currently doing this? Couldn’t we just take the ffmpeg from those builds, break it apart and figure out how its being enabled and put that into a custom build of plex for x86/x64?
Wait… So plex is currently doing this? Couldn’t we just take the ffmpeg from those builds, break it apart and figure out how its being enabled and put that into a custom build of plex for x86/x64?
My guess is that nVidia assisted them in making the Shield client work. I wouldn’t be surprised even if nVidia approached them. But it might not be as simple as just moving the transcoder, especially from the Shield given that it’s Android and we don’t know what was done to make it work.
@peva said:
I wonder if Plex Cloud is leveraging some sort of hardware transcoding? Just a random thought.
I just assumed it was using the Amazon Elastic Encoder, which has been around for a long time. And who knows what Amazon is doing on the backend for that. Maybe custom code running on a CUDA backed farm or something. There’s some performance/cost-effectiveness there that I could see making sense.
There again, the ongoing cost associated with that would suggest that Plex would have to have some kind of agreement with Amazon to make it financially viable.
@MasterChiefmas said:
I just assumed it was using the Amazon Elastic Encoder, which has been around for a long time. And who knows what Amazon is doing on the backend for that. Maybe custom code running on a CUDA backed farm or something. There’s some performance/cost-effectiveness there that I could see making sense.
There again, the ongoing cost associated with that would suggest that Plex would have to have some kind of agreement with Amazon to make it financially viable.
I am also very interested in how Plex Cloud will handle transcoding. Maybe it could be applied outside the Amazon too. I know there are some attempts at remote transcoders but it would be nice to have something of a “turnkey” solution. Without all the headaches .
@peva said:
I wonder if Plex Cloud is leveraging some sort of hardware transcoding? Just a random thought.
I just assumed it was using the Amazon Elastic Encoder, which has been around for a long time. And who knows what Amazon is doing on the backend for that. Maybe custom code running on a CUDA backed farm or something. There’s some performance/cost-effectiveness there that I could see making sense.
There again, the ongoing cost associated with that would suggest that Plex would have to have some kind of agreement with Amazon to make it financially viable.
Can’t find the specific post but a Plex Employee said that they aren’t using AEE. I hope they tell us more about Plex Cloud when it’s out of Beta
It would be nice if they would implement it, it was only requested 4 years ago. Funny how they spend so much time and effort on the shield when its probably the smallest market out of any server yet they cant commit to quick sync which would be used by probably 90% of people. Their priorities are mind boggling.
@nagle3092 said:
It would be nice if they would implement it, it was only requested 4 years ago. Funny how they spend so much time and effort on the shield when its probably the smallest market out of any server yet they cant commit to quick sync which would be used by probably 90% of people. Their priorities are mind boggling.
True however expanding market is always nice to get new clients, which might not be so tech savvy. I do agree that Quick Sync should come sooner rather than later. On another note those of us who has support for Quick Sync does have some pretty powerful CPUs so the need for quick sync is not that pressing.
@nagle3092 said:
It would be nice if they would implement it, it was only requested 4 years ago. Funny how they spend so much time and effort on the shield when its probably the smallest market out of any server yet they cant commit to quick sync which would be used by probably 90% of people. Their priorities are mind boggling.
True however expanding market is always nice to get new clients, which might not be so tech savvy. I do agree that Quick Sync should come sooner rather than later. On another note those of us who has support for Quick Sync does have some pretty powerful CPUs so the need for quick sync is not that pressing.
Not necessarily. QuickSync was introduced with Sandy Bridge, so my 2011 Mac mini (and I believe my 2011 MacBook air) supports it… Yet it has a passmark barely above 2500. Would be very useful… Though id have to run everything off windows in bootcamp.
Can’t find the specific post but a Plex Employee said that they aren’t using AEE. I hope they tell us more about Plex Cloud when it’s out of Beta
That would be very interesting if true. I had assumed a hardware encoder somewhere, it just doesn’t seem like it’d be cost effective if they were using just straight general purpose CPUs and software encoders, but like I said, I don’t know how AEE does it. Maybe they have some new thing that AWS gave them access to that allows them to access encoder blocks on the hardware and it’s just not classified as part of AEE yet.
Can’t find the specific post but a Plex Employee said that they aren’t using AEE. I hope they tell us more about Plex Cloud when it’s out of Beta
This is pure speculation, but it’s possibly not entirely coincidental that Plex Cloud was announced right around the same time that Amazon introduced their new P2 EC2 instance types, which are specifically geared towards GPU workloads:
“This new instance type incorporates up to 8 NVIDIA Tesla K80 Accelerators, each running a pair of NVIDIA GK210 GPUs. Each GPU provides 12 GiB of memory (accessible via 240 GB/second of memory bandwidth), and 2,496 parallel processing cores.”
They could be using the same hardware transcoding engine from the NVIDIA Shield to offload processing to Amazon’s NVIDIA GPUs.
Can’t find the specific post but a Plex Employee said that they aren’t using AEE. I hope they tell us more about Plex Cloud when it’s out of Beta
This is pure speculation, but it’s possibly not entirely coincidental that Plex Cloud was announced right around the same time that Amazon introduced their new P2 EC2 instance types, which are specifically geared towards GPU workloads:
“This new instance type incorporates up to 8 NVIDIA Tesla K80 Accelerators, each running a pair of NVIDIA GK210 GPUs. Each GPU provides 12 GiB of memory (accessible via 240 GB/second of memory bandwidth), and 2,496 parallel processing cores.”
They could be using the same hardware transcoding engine from the NVIDIA Shield to offload processing to Amazon’s NVIDIA GPUs.
It’s been said clearly that transcoding isn’t done on Amazon’s side of things, it’s done on Microsoft Azure.
It’s been said clearly that transcoding isn’t done on Amazon’s side of things, it’s done on Microsoft Azure.
It’s been said clearly that transcoding isn’t done on Amazon Elastic Transcoder (which is totally different than Amazon EC2), but AFAIK nothing has been said otherwise. Where did you read that it’s done on Microsoft Azure? I’m not saying you’re wrong, but that doesn’t really make any sense at all and I find that info quite hard to believe.
.