mirror of
https://github.com/jellyfin/jellyfin.git
synced 2026-05-04 18:09:12 +03:00
Setting too high of a bandwidth for ffmpeg trips up NVENC (and probably other hardware encoders) #802
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @EraYaN on GitHub (Jul 22, 2019).
Describe the bug
Reference: https://www.reddit.com/r/jellyfin/comments/cgac40/jellyfin_wont_use_hardware_transcoding_on_some/eufpgp6 (and childs)
So when jellyfin sets a bandwidth too high for the encoder the encoder errors out.
So
-b:v 329027765is clearly to high for this hardware (GTX650)To Reproduce
-b:vargument outside the acceptable ranges for the encoder.Expected behavior
Jellyfin should cap these given a table of encoders and their max bitrates.
Logs
(See reddit post)
System (please complete the following information):
Additional context
This will hold for quicksync/vaapi/omx and others too I presume. NVidia does not publish these limits it seems.
EDIT: My GTX1080Ti seems to top out at 144M for the bitrate. I sense reason in the original 140MBps number ;)
@k1lln1n3 commented on GitHub (Jul 30, 2019):
Is this something we also see cumulatively? Like will multiple streams adding up to a high bitrate have the same behavior? I was going to submit a feature request for the ability to determine a limit of streams sent to hardware encoders. Wondering if it could be linked back to this as well.
@stale[bot] commented on GitHub (Oct 28, 2019):
Issues go stale after 90d of inactivity. Mark the issue as fresh by adding a comment or commit. Stale issues close after an additional 14d of inactivity. If this issue is safe to close now please do so. If you have any questions you can reach us on Matrix or Social Media.
@EraYaN commented on GitHub (Oct 30, 2019):
bot go away
@sevenrats commented on GitHub (Oct 23, 2023):
@EraYaN @k1lln1n3 a test file that reproduces this error would make a huge difference in getting it resolved.
@EraYaN commented on GitHub (Oct 24, 2023):
Ooh this is file independent, it's just the maximum output bandwidth. You can set anything as the input frankly. I think modern NVENC caps out at like 1200Mbits or something? But older cards cap out much much lower (as described above) and it's depends on the chosen codec it seems. I'm not at my home PC to try it out but ffmpeg craps out immediately so getting the limits is not that hard if you have the hardware.
@k1lln1n3 commented on GitHub (Oct 24, 2023):
@EraYaN is correct. This was a collective amount of data thrown at the older 10 series cards. I believe I Was able to make it happen with 2 separate files I had at the time but I couldn't tell you what they were. I''m not even sure this is an issue anymore as that card is long gone.