[Drag0nFly]

I'm not using mainline. All my tests were done with the most recent Xvba build from your PPA.

[Drag0nFly]

btw.–maybe it is better to ask this question in the main Xvba thread? I put it here since it specifically dealt with Intel GPUs, while the other thread seems to focus more heavily on ATI & Nvidia.

[wsnipex]

unless you already run the latest version, upgrade, try again and post in the other thread if the issue persists.

[ncube2]

delete

[bwat47]

For ref - the full range/limited colour range patches are set for the 3.9 kernel. By default, unless something changes it should default to automatic - for CEA defined "tv modes" this will be limited range, for all other modes full range. You can still force the setting using xrandr. Unfortunately it seems the automatic selection "fails" on non-integer framerates (e.g. 23.976, 59.94) and that's just how it is for now.

For this (and for also disabling my laptops display and using the hdmi display if its connected) I've configured lightdm to run a script with xrandr settings on boot by editing /etc/lightdm/lightdm.conf and adding display-setup-script=/path/to/your/script.sh. seems to work well so far.

and for the script:

Code:

EXTERNAL_OUTPUT="HDMI1"
INTERNAL_OUTPUT="LVDS1"

xrandr | grep $EXTERNAL_OUTPUT | grep "connected"
if [ $? -eq 0 ]; then
    xrandr --output $INTERNAL_OUTPUT --off --output $EXTERNAL_OUTPUT --auto --output $EXTERNAL_OUTPUT --set "Broadcast RGB" "Limited 16:235"
else
    xrandr --output $INTERNAL_OUTPUT --auto --output $EXTERNAL_OUTPUT --off
fi

[costi]

The last update to the guide was 1.5 years ago (at least according to the second post). Has anything significant changed since then?

[Drag0nFly]

Apologies ahead of time as this thread appears to be dead; but since I've spent the last few days googling for this issue (which always came up inconclusive), I might as well post it here:

Has anyone successfully set up their Ivy Bridge GPU (specifically, the Intel HD4000) to output proper (or close to) 23.976 with XBMC? The closest I get is 24Hz (according to XBMC's status screen, and "23.99", according to my projector), which causes a skipped frame about once every minute.

I know there has been a lot of back-and-forth wrt. Intel supporting this (was broken on Sandy Bridge, supposed to be fixed on IB on all but one Intel board (DH77EB), "definitely" fixed on Haswell, etc.), but the reports on the Intel forums, Anandtech, etc., are all pretty much inconclusive.

What has been tried -

- Updating the drivers from Intel Open Source Technology Center (01.org) to the latest versions (i965, vaapi, libva)
- Updating the BIOS (I am on a Shuttle XH61V)
- Playing with X modelines and xrandr

I assume there isn't much XBMC can do to force proper 23.976 output as long as the hardware doesn't properly support it. But if anyone has similar experiences or pointers I would very much like to hear them.

This is on a Ubuntu 13.04 (Raring) system running a 3.8.0-35-generic kernel, btw.

[fritsch]

Upgrade your kernel to something >= 3.11.3 and see ...

[Drag0nFly]

Yes, I was planning on doing that as soon as the installation is completely backed up. However, I couldn't find anything in the changelogs which would suggest the kernel having an impact on the refresh rate.

What happened in 3.11.3, btw.? (The newest official kernel for saucy is 3.11.0-17)

[fritsch]

Intel implemented fractional refreshrate detection. I just named the "oldest version". 3.13.5 is also fine.

[pumkinut]

23.976Hz material is not fixed on Ivy Bridge. IB suffers the same fractional framerate hardware bug as Sandy Bridge. There's no driver or BIOS update that will fix it, it's a hardware level issue. It's fixed in the Haswell based processor line.

[wsnipex]

well, mine at least shows a perfectly stable 23,98 hz in the codec debug overlay.

[Drag0nFly]

Thanks for the input guys-I will hopefully be able to try this later today (most likely with the 3.11.10.4 build from http://kernel.ubuntu.com/~kernel-ppa/mainline/)
Then after a practical test I can confirm or deny whether this had an impact (at least on this Shuttle system with the HD4000 iGPU, running off a i3-3225 3.3Ghz CPU)

[Drag0nFly]

Well, this was a pleasant surprise–upgrading the kernel certainly did make an impact on this issue. I waited a while to post in order to verify that this wasn't simply spurious behaviour, but actually a real improvement from what I've witnessed earlier. Kudos to Intel for adding fractional refreshrate detection and the XBMC team for suggesting to try it out.

I've watched two movies in full, as well as a couple of episode series, and can note that if the problem isn't completely eliminated (still too early to tell yet), its presence has been significantly reduced.

For the record, I ended up upgrading the kernel to 3.13.5 (while keeping everything else the same, including an old XBMC Xvba build), and could immediately see the change in the picture fluididty when watching for ~5 mins.

The logs also report a display resolution adjustment to 23.98Hz (vs. 24Hz previously) –

grep ADJUST .xbmc-27-Feb-2014/temp/xbmc.log
21:13:26 T:140370268649216 NOTICE: Display resolution ADJUST : HDMI2: 1920x1080 @ 24.00Hz (19) (weight: 0.001)
21:25:56 T:140370251863808 NOTICE: Display resolution ADJUST : HDMI2: 1920x1080 @ 24.00Hz (19) (weight: 0.001)
20:54:51 T:140370226685696 NOTICE: Display resolution ADJUST : HDMI2: 1920x1080 @ 24.00Hz (19) (weight: 0.001)

grep ADJUST .xbmc/temp/xbmc.log
21:44:56 T:140178194687744 NOTICE: Display resolution ADJUST : HDMI2: 1920x1080 @ 23.98Hz (21) (weight: 0.000)
21:48:11 T:140178203080448 NOTICE: Display resolution ADJUST : HDMI2: 1920x1080 @ 23.98Hz (21) (weight: 0.000)
21:49:23 T:140178194687744 NOTICE: Display resolution ADJUST : HDMI2: 1920x1080 @ 23.98Hz (21) (weight: 0.000)

For some reason though I couldn't get accurate readings from the overlayed XBMC status screen (the one activated when pressing 'o')–it shows a constant 24.00fps.
I also upgraded to the latest version offered by FernetMenta's branch, but it remained the same. Wonder if it has to do with old config files in .xbmc

Anyway–I am very happy to have this working on my "non-recommended" Compositing setup (Cinnamon), running a separate desktop in addition to XBMC. I will do more testing as I am currently migrating all my Blu-Ray's to the media server (*goodbye* BD-Java, FBI/Homeland/Interpol warnings, unskippable trailers + resume at end-credits _forever_)

I guess the remaining issues are 50Hz + 3D playback via XBMC. But knowing the team I am sure those will mature soon as well. :-)

[Drag0nFly]

An update–

Everything is still working quite smoothly (pun intended after switching to the 3.13.5 kernel. There is still the occasional frame drop, but it is occurs much less frequently than before. Video startup (initial resolution adjustment) also appears to be a little faster, but that is probably more as a result of running a newer XBMC build.

I also notice that there has been huge improvements to the way VC-1 material is handled when VAAPI acceleration is enabled. No more segfaults (*knocks wood*) when trying to play material using this codec. Cool. CPU usage when viewing a full-quality, high-bitrate Blu-Ray rip even during intense action scenes is less than 10% usage for each core on my i3-3225. Impressive to say the least!

Regarding the codec debug overlay output–this appears to be related (on my system at least) to the 'Adjust display refresh rate to match video' setting when set to 'On start/stop'. When set to 'Always' it shows 23.98, but the video is not smooth. Anyway–this is cosmetic.

The only issue I've encountered so far is actually related to the audio engine and output to optical (TOSlink). The newer XBMC builds exhibit a half-second audio stutter when pausing the video, presumably to clear the audio buffer. When resuming the audio is ok, but the error is reproducible every time a movie is paused.
My primary testing was done with DTS HD MA material (where the DTS Core is output through the optical port), but I assume it occurs regardless of audio codec. (I'll be able to verify that hopefully later today.)

I did do some tests on 50Hz / 25 fps (stutter) & 3D/MVC material as well (unsuccessfully); but since the information regarding support for 3D in XBMC is scarce I assume the support isn't ready yet. Proper 50Hz output is probably also a harder nut to crack...