Closed johnwchambers closed 4 years ago
That looks correct. The hwaccel option reads what's supported by FFMPEG, nothing is hard coded so there shouldn't be any limitations there You do also need to use the accompanying nevc codec to get the advantages of hwaccel for encoding by setting your codec to h265_nvenc or h264_nvenc
I'm not 100% sure the difference between using cuvid vs cuda though, but it looks like cuda doesn't have a decoder
awesome. i updated my autoprocess ini with the below. those look like the ones that were brought in from compiling with the nvidia drivers. although... i don't know what the difference between each is lol. looking at each of their capabilities looks exactly the same with "ffmpeg -h encoder=xxx".
codec = h264_nvenc, nvenc, nvenc_h264
Thanks!
nvenc, nvenc_h264 aren’t valid so you can leave those off. Include h264 by itself too so that you can still remux
Roger that, thanks again!
Curious about your experience with the P2000 card remuxing your video files with ffmpeg. I was looking at this card, but wasn't really able to justify spending $5-600 for it. I had a GTX-1660 card bought for another project. I built a new Linux box around this card with dedicated purpose of converting video. Once I was able to compile a working ffmpeg build with nvenc support. And use SMA, my remuxing times (45min 1080p TV Show to 720p) dropped from 21 mins (software) to 2.5 mins (hardware). Took some tweaking and help from @mdhiggins, but Its been an amazing difference and worth the effort.
If I was to get a P2000 it would be to implement the parallels transcoding thats available in a high end card. I can do it with my graphics card, but not really getting any benefit for the extra scripting required to pull it off.
When I purchased the P2000 I bought it along with a brand new physical server so that I could offload plex and and all the associated services (sonarr/radarr/bazarr/jackett/nzbget/qbittorrent/nzbhydra2/ombi/sma/plex-autoscan) with it off of my virtual lab equipment to get back all the CPU/RAM resources I was giving them as multiple VMs. With that said, because of the new physical server combined with the P2000, my satisfaction with my new setup went up 100 fold.
My main reason for getting the P2000 card was due to me having 20+ users on my server and hardly any of them updating their playback settings lol. But I figured since it supports hardware transcoding, might as well compile ffmpeg with the capability and configure it within sma to utilize the card as much as possible. Thus far I'm extremely satisfied with it and now don't really care who is transcoding versus direct playing my media.
I had a customers new computer (Xeon, 32gb RAM & P2000) that I had access to for a couple days before I had to deliver. I tried to build out a Plex server with hardware transcoding. But I never got the build to work before I had to deliver.
My current Plex build is around i7-4790K, 32Gb and SSD Drives. I typically have 4-6 remote users, but have up to 8 concurrently. My bottleneck is the available outbound bandwidth. I have 30mb upload speeds. I don't "pre" transcode any of my movies, only the tv shows. Two things I did that seemed to help a lot was create a RAM drive that's used by Plex to transcode on the fly. And I use bonded 2 x 1Gbit network connections between all my servers, storage, switches and router.
Awesome sauce. If you ever wanna geek out about build differences, scripts, or etc, check out my github profile page for my email addy.
I just acquired a new physical server and installed an NVIDIA Quadro P2000 in it. I've compiled ffmpeg with nvenc support which shows cuda as my only available hardware acceleration. Is cuda supported as an option in the autoProcess.ini for hwaccel? Or am I missing something that I should have added to the build? Operating system is on CentOS 7.
Thanks!
ffmpeg build info
ffmpeg hwaccel output:
ffmpeg encoders output:
ffmpeg decoders output: