[x86-64][LibreELEC x CB1 Builds] Dolby Vision FEL, HDR10+ and Full 3D

  • Hi ashlar,

    Thanks for your interest! 😃

    Unfortunately, my Optoma UHD50 projector does not support Dolby Vision, so I can't properly test Dolby Vision FEL.

    However, I can confirm that 4K UHD playback works beautifully on my HP EliteDesk 705 G4 SFF with the new LibreELEC 13 build.

    I tested King Kong in 4K yesterday, and the picture quality on my Optoma projector was absolutely stunning!

    The projector correctly recognizes the UHD resolution. Although it supports HDR10, I haven't verified whether HDR10 was actually active during playback.

    Hopefully, someone with a Dolby Vision-capable display can test FEL on the same hardware.

  • if your cpu is amd 2400g, then it has the same igpu what i have (cpu 3400g).

    it can play dv fel !!!

  • if your cpu is amd 2400g, then it has the same igpu what i have (cpu 3400g).

    it can play dv fel !!!

    It's not a question if the Chopset can play FEL, it's if the fact he's using a module for HDMI, which converts DP to HDMI, will it work over the module is what we'd like to know

  • Using projectors mainly... an XGIMI Titan Noir Max and an XGIMI Aura 3 Max. Settings for these builds mostly default, other than for 3D w/Preferred and Same as Movie, which I show in my YT videos.

  • CroqueMr Since you mentioned that you prefer your issues and logs over on github i created an issue over there with a log attached. Might not have seen your answer to my first message, so if my hardware is not of interest for you to consider supporting i sincerely apologise for the additional clutter.

  • Hi CroqueMr,

    I tested the new release on my **Intel Arc B570** setup (connected via Denon AVR-X2800H to Sony A80K OLED):

    1. **Profile 7 FEL Mid-Stream Resume Causing Continuous Frame Skips:**
    - Starting a Profile 7 FEL movie from `00:00:00` plays smoothly with `0` dropped and `0` skipped frames.
    - However, selecting **Resume** mid-movie causes playback to stutter every alternate second, and the Player Info (`O`) overlay confirms it is rapidly accumulating **Skipped** frames (not Dropped frames), indicating an A/V clock / PTS desync between the dual-layer VAAPI decoders and the audio stream upon mid-file re-entry.
    - Confirmed this happens identically on both **local files** and network streams.
    - Debug log reproducing the local FEL resume frame-skipping:

    https://paste.libreelec.tv/moralless-alycia.log



    2. **HDR10 & Video Stop:**
    - [Add note here: e.g., HDR10 playback and stopping video back to GUI no longer crash/hang]

  • CroqueMr Since you mentioned that you prefer your issues and logs over on github i created an issue over there with a log attached. Might not have seen your answer to my first message, so if my hardware is not of interest for you to consider supporting i sincerely apologise for the additional clutter.

    perfect thanks, do not apologise


    ty I will check that

  • Hi CroqueMr,

    Confirmed one critical detail: **any HTTP range request (both `Resume` and seeking forward during playback)** causes `CFileCache` to collapse from `45%` straight to `0%` (`Buffers V/A` dropping to `54/-6 % : 72/14 %` and `DROP / SKIP` climbing again).

    - **Local disk (`Videos -> Files`):** Both seeking and Kodi `Resume` keep `0 / 0` drops/skips.
    - **Debrid HTTPS stream (MegaNexus):** Starting at `0:00` fills `PLAYER CACHE` to `45%` with `0 / 0` drops/skips, **until** you seek forward or use `Resume`—at which point the HTTP range re-entry drops `PLAYER CACHE` permanently to `0%`.

    So the bug is specifically how Kodi's `CFileCache` / Curl handles HTTP `Range` requests on re-entry/seek in this build.


    One more confirmation: I just tested a **Dolby Vision Profile 7 MEL** stream (`HDR10 BL + MEL + RPU`, `35.4 Mb/s` + `3.2 Mb/s` TrueHD Atmos, `FEL / upscaling: No FEL`) over the exact same Real-Debrid HTTPS connection:

    - **`PLAYER CACHE`:** `99%`
    - **`DROP / SKIP`:** `0 / 0` (even after seeking/resuming)
    - **`Buffers V/A`:** `76/-7 % : 98/18 %`

    So the `0%` `CFileCache` collapse on HTTP range requests (`Seek` / `Resume`) happens **exclusively when the dual-layer `Full FEL - Reference GPU` (`vaapi`) pipeline is active** (simultaneous BL + EL reads over HTTPS/WebDAV), whereas single-layer MEL over HTTPS and dual-layer FEL from local disk both run at `0 / 0` drops/skips.

    https://paste.libreelec.tv/humeral-jefferey.log

    Edited once, last by lvasan: Merged a post created by lvasan into this post. (November 13, 2026 at 5:48 PM).

  • Im tested both RC3 and RC4 preview in as much time as i had for now with ASUS NUC 14 Pro s Core Ultra 5 125H with Intel Arc GPU. Transformers, Tron and other mentioned work fine with Full FEL7 decoding, no skipped frames, nothing, it just works great.

    What i did notice is that after the playback is stopped audio output device is lost and there is no more audio. For example i have audio output set to ALSA: HDA Intel PCH, GSM LG TV SSCR2 on HDMI #0.

    When i stop the playback i only have ALSA: HDA Intel PCH, HDMI #0.

    When i reboot i again have LG HDMI connection. I tried several combination, it doesn't mater which video or audio codec was used before stopping the playback.

    Also, after some 30 minutes to an hour GUI losses 60fps and goes sluggish, probably 30fps or even lower. After reboot its again 60fps. Its the same when GUI is set to 4K or 1080p.

  • I'm the Tiger Lake-H laptop user from my earlier report (Kodi main thread blocking in `drm_syncobj_wait_ioctl`). After updating to **RC4 Preview Update 4** (build `3d185e8381df3234982e13003f747b49390153bf`), I can no longer reproduce the freezes: Bad Apples (SDR), The Odyssey (HDR10), and three Dolby Vision / HDR10 titles all play in their native format, without having to force the 8-bit bit-depth option. Thank you for the work on this.

    One problem remains on every film I tried: **pressing O always causes a burst of dropped and skipped frames** (about one second of video), then playback is clean again. It does not depend on bitrate or on cache fill.

    ## System
    - LibreELEC (community) DV-R1.0.0-RC4-Preview-Update4, Kodi 22.0-RC1 (21.90.901), kernel 7.2.6
    - ASUS ROG Zephyrus M16 (GU603HM): i7-11800H, Intel UHD Graphics (TigerLake-H GT1) drives HDMI-A-1; RTX 3060 Mobile has no connected output
    - Output: HDMI to an Onkyo AV receiver, then to a TV; 3840x2160 @ 60 Hz, HDR/DV capable EDID
    - Boot: NVMe SSD, `video=eDP-1:d intel_iommu=igfx_off`
    - Content played over SMB / Plex direct play on a wired 1 Gbps network
    - Settings: Dolby Vision playback = 4 (HDR10 Rendering); HDR10 playback = Kodi default; "Enhanced player information" = on (default)

    ## Frame drops when pressing O (10-second window containing the press)

    | Film | Format | fps | Bitrate | Dropped / skipped | video_queue_pct |
    |---|---|---|---|---|---|
    | Gemini Man | Dolby Vision | 59.94 | ~85 Mbps | **105** / 4 | 99–100 |
    | Primetime | DV / HDR10+ WEB-DL | 23.976 | ~14 Mbps | 29 / 1 | 95 |
    | Justice League | HDR10 | 23.976 | ~51 Mbps | 18 / 2 | 97 |
    | The Odyssey | HDR10 | 23.976 | ~89 Mbps | 19 / 2 | 94 |
    | Bad Apples | SDR | 23.976 | n/a | no DVBridge statistics on this path | n/a |

    Observations:
    - In every case the first sign is `OutputPicture - timeout waiting for buffer` within ~0.5 s of `Loading skin file: special://xbmc/media/cb1/DialogPlayerProcessInfo.xml`. After the burst, drop_delta returns to 0.
    - `video_queue_pct` stays at 94–100% (about 3.7–4 s of video queued) throughout, so the source and decode side keep up. The stall looks like it is on the output/presentation side when the panel is composited over a directly scanned-out video plane.
    - Bitrate does not matter: the 14 Mbps file dropped more frames than the 89 Mbps one. At 59.94 fps the same stall costs more frames (105), which fits a fixed ~1.7 s stall.
    - Earlier in the same session, `DVBridge renderer summary` shows `direct=N composed=0`, so the video is on the direct path until the panel opens.

    Attached logs for reference. Thanks you so much for your work!

  • Hi CroqueMr,

    Thanks again for all your work on libreelec-x86-DV. The Dolby Vision picture quality and punch on this build is much better than CoreELEC, so I really want to keep using it as my main setup.

    However, even with 64 MB cache, Buffer all filesystems, Adaptive read factor, and 128 KB chunk size, Dark Shadows (2012) (88 Mb/s Profile 7 FEL) still drops to 0% PLAYER CACHE and hits 1 / 152 skipped frames in under two minutes while CPU is at 2%. Profile 7 MEL plays at 99% cache with 0 / 0 drops/skips.

    Here is the debug log from that 0% cache / 152 skip run:
    https://paste.libreelec.tv/unmoralizing-nina.log

    Since this happens across WebDAV and local playback whenever Full FEL - Reference GPU is active, it looks like a CFileCache / dual-layer read conflict in the build. Could you please take a look when you get a chance? Happy to test any build or fix right away.

  • Im tested both RC3 and RC4 preview in as much time as i had for now with ASUS NUC 14 Pro s Core Ultra 5 125H with Intel Arc GPU. Transformers, Tron and other mentioned work fine with Full FEL7 decoding, no skipped frames, nothing, it just works great.

    What i did notice is that after the playback is stopped audio output device is lost and there is no more audio. For example i have audio output set to ALSA: HDA Intel PCH, GSM LG TV SSCR2 on HDMI #0.

    When i stop the playback i only have ALSA: HDA Intel PCH, HDMI #0.

    When i reboot i again have LG HDMI connection. I tried several combination, it doesn't mater which video or audio codec was used before stopping the playback.

    Also, after some 30 minutes to an hour GUI losses 60fps and goes sluggish, probably 30fps or even lower. After reboot its again 60fps. Its the same when GUI is set to 4K or 1080p.

    Could you share the log please.

  • Hi CroqueMr,

    Thanks again for all your work on libreelec-x86-DV. The Dolby Vision picture quality and punch on this build is much better than CoreELEC, so I really want to keep using it as my main setup.

    However, even with 64 MB cache, Buffer all filesystems, Adaptive read factor, and 128 KB chunk size, Dark Shadows (2012) (88 Mb/s Profile 7 FEL) still drops to 0% PLAYER CACHE and hits 1 / 152 skipped frames in under two minutes while CPU is at 2%. Profile 7 MEL plays at 99% cache with 0 / 0 drops/skips.

    Here is the debug log from that 0% cache / 152 skip run:
    https://paste.libreelec.tv/unmoralizing-nina.log

    Since this happens across WebDAV and local playback whenever Full FEL - Reference GPU is active, it looks like a CFileCache / dual-layer read conflict in the build. Could you please take a look when you get a chance? Happy to test any build or fix right away.

    I will try to reproduce the issue it's strange, if I can not manage to reproduce it I think we will reopen the conversation on the support build ;)


    Dropping frame with geminiman is normal, your GPU is overloaded. But for the other no, I can not manage to reproduce it, I will adjust the log for the next release

  • I will try to reproduce the issue it's strange, if I can not manage to reproduce it I think we will reopen the conversation on the support build ;)

    Hi CroqueMr,

    Sounds good, thank you for looking into it! If you aren't able to reproduce it on your setup, you are more than welcome to connect directly to my system over SSH (or whichever remote access method you prefer) to inspect and test it live on my Intel Arc B570 setup.

    Just let me know whenever you are ready and I will send you the SSH details via private message.

  • Hi CroqueMr,

    Sounds good, thank you for looking into it! If you aren't able to reproduce it on your setup, you are more than welcome to connect directly to my system over SSH (or whichever remote access method you prefer) to inspect and test it live on my Intel Arc B570 setup.

    Just let me know whenever you are ready and I will send you the SSH details via private message.

    I am finalizing the next update and I will compile the remote acces build. I will ping you tomorrow for the tests on the private repo