Slow hardware acceleration on ds418play for 1080p

Server Version#: 1.14.0.5470 (64 bit)
Player Version#: 3.72.1

Trying to play a 1080p h264 video on a samsung tv, with plex media server running on a synology ds418play. It was going quite slow, buffering about once every 30s.
I’ve now bought a plex pass, to enable hardware acceleration, hoping it would improve, but it did not.
Hardware acceleration seems to be enabled (I can see the ‘(hw)’ in the now playing screen), but if I go through the logs I still see that the transcoding speed averages at about 0.8. My synology cpu is still being hammered at about 70% with or without hardware acceleration enabled. I’ve restarted my synology as well without any effect.

As far as I can find (compatibility list and so) ds418play should be able to cope with 1080p no problem when using hardware acceleration.
Am I missing something?

I can send logs if need be.

Are you using subtitles?

Yes I am.
Disabling them does seem to improve the speed to 1.2 on average. CPU usage still at roughly 70%.

So using subtitles is problematic?

Edit: tried without subtitles and with hardware acceleration disabled, then it averages at about 1.0
So hardware acceleration does seem to do something, but a lot below what I was expecting.

Yes.

The ASIC in the CPU provides the video processing.
Subtitles must be overlayed by the CPU. The ASIC doesn’t provide that functionality else we would have taken advantage already

I would like to see the XML (Get info -. View XML) of the item you’re playing.
Also, knowing the player settings (is it limiting anything ?) helps.

Lastly,

  1. Debug level logging only
  2. Start Playback
  3. Let it run 30 seconds into the movie/episode
  4. Stop
  5. Wait 30 seconds to clean up
  6. Settings _ Server - Troubleshooting - Download Logs to gather the logs
  7. Attach the ZIP it gives you.

I’ve sent the xml and log in pm.

Player settings:
Local/remote/online quality: original
burn subs: auto
Allow direct play: on
Allow Direct Stream: on
max h264 level: 5.1

Thank you for turning off Verbose. it wastes more than it provides at this level.

1… Your database is fragmented badly or your system is grossly low on memory and swapping. 78 seconds indicates memory swapping.

Dec 04, 2018 23:32:10.179 [0x7f3288047700] DEBUG - It took 90.000000 ms to retrieve 2 items.
Dec 04, 2018 23:32:10.179 [0x7f327fbff700] DEBUG - It took 80.000000 ms to retrieve 2 items.
Dec 04, 2018 23:32:10.198 [0x7f328748f700] WARN - SLOW QUERY: It took 78530.000000 ms to retrieve 50 items.
Dec 04, 2018 23:32:10.204 [0x7f328777d700] WARN - SLOW QUERY: It took 760.000000 ms to retrieve 50 items.
Dec 04, 2018 23:32:10.363 [0x7f328748f700] DEBUG - Setting container serialization range to [0, 6] (total=-1)
Dec 04, 2018 23:32:10.368 [0x7f328777d700] DEBUG - Setting container serialization range to [0, 6] (total=-1)
Dec 04, 2018 23:32:10.374 [0x7f328777d700] DEBUG - There were 1 top-level paths for Mobile Suit Gundam 00.

Do you have more than 2GB installed ? 8GB is optimal (and max physical) for the NAS.

No, the standard 2GB RAM. Memory swapping would surprise me, given it’s at about 50% utilization.
Plex is about the only memory-heavy process, running at about 270MB.

As for my drives, if it matters, those are 4 WD Red in SHR raid.

Actually, PMS (as a total system) takes more than 270 MB.
What you’re also not seeing is the memory used by DSM itself. (been adjusted out of what you see by the UI). PMS does utilize more than 270 MB. Heap allocations and other buffers don’t show in totals.

I do strongly recommend 4GB of RAM. You will see a profound improvement.
As others will attest to, 8GB , while maximum, is also a sweet spot. HW transcoding of two 2160p images to 1080p fits very nicely in that memory space (those are big images)

Huh, good to know. I’ll take a look for that RAM upgrade.
Looking at resource monitor of synology, the swap section, it shows indeed some swapping. Reasonably small, but swapping nonetheless.
12% used of 3gig swap, with in/out spikes of up to 40KB/s. Not sure whether this is sufficient to explain the slow query.

it is. there should be no swapping. PMS uses semaphore locks on the database.
If that block of code gets swapped out then , no matter how small, bad news.

Aha, makes sense, thanks!
Price for 2x4GB RAM seems reasonable, so I’ll probably go for the upgrade, even if it would end up not improving the situation.

a matched pair, 8 GB “kit” guarantees you the best performance.

Uhu, I know, thanks!
Just ordered a pair.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.