[fritsch]

Suspend and Resume is optional - whenever the AVR is not yet up, you will get problems, a custom xorg.conf or a custom EDID will help.
I updated the kernel to 3.18.1 see first post.

[klmd]

Kernel 3.18.1+: Anyone else having an issue with the system fan running full speed for like 20 seconds upon resume from standby?

ASUS Chromebox running Mint 17, didn't happen with 3.16. Happens with 3.18.1 and 3.19 RC1.

I'd like the fixes that come with the kernel update but the fan is really loud for those 20 seconds until it stops. The workaround I suppose would be shutting down every time instead of entering standby,

[Sta11ion]

Encountered GPU HANG on Haswell Refresh + Z97 on stock 3.16 kernel. Going to install new kernel...

BTW, just noticed Chris Wilson's patches have been merged in 3.19: http://kernel.ubuntu.com/~kernel-ppa/mai...id/CHANGES

EDIT: 3.19 kernel .deb are here: http://kernel.ubuntu.com/~kernel-ppa/mai...rc2-vivid/ Haven't tested it on Trusty, should work

[fritsch]

Jep, I know - but don't want users to run into instabilities (RC kernels) cause of that. I always get a mail when that patch is pushed somewhere :-)

[tavoc]

I just wanted to say thanks @fritsch (do you have a paypal donate account? ) . Tested your modification on my Xbmcbuntu setup with kodi (stable).
The deinterlacing of SD Live Content has greatly improved.



You stated that "If you are running a Baytrail NUC or a 847 Celeron NUC or comparable - choose Scaling: Bilinear and Deinterlace-Method: BOB".
I have a j1900 (asrock q1900 tm board with 8 gb) and can use Deinterlace: Auto and Deinterlace Method : Deinterlace and lanczos3 with no problems.

Is there something why I should stay with your recommend settings?

[fritsch]

I just wanted to say thanks @fritsch (do you have a paypal donate account? ) . Tested your modification on my Xbmcbuntu setup with kodi (stable).
The deinterlacing of SD Live Content has greatly improved.



You stated that "If you are running a Baytrail NUC or a 847 Celeron NUC or comparable - choose Scaling: Bilinear and Deinterlace-Method: BOB".
I have a j1900 (asrock q1900 tm board with 8 gb) and can use Deinterlace: Auto and Deinterlace Method : Deinterlace and lanczos3 with no problems.

Is there something why I should stay with your recommend settings?

Nice you really get Lanczos3 Optimized going with the 576i to 1080p without any drops? Last time I tested that was not possible - at least not with my J1800.

Concerning the donation: You can donate for xbmc - http://kodi.tv/contribute/donate/ Thank you very much in the name of the team.

[tavoc]

In SD Channels (RTL, RTL2, Pro7 and so on) and HD Channels (I think only 720p or 1080i. Over DVB-S2 in Germany are no channels with 1080p?) it runs fine with the aforementioned settings. I see a lot of skipped frames (as before without your modification), but very few dropped frames. Only in the first few seconds there is a drop (but it does not increase over the time).


If you like I can take a screenshot of my settings and the resulting image (with detailed overlay information).



Thanks for your donationlink.

[fritsch]

The "skipped" frames is the render much too slow, that's caused by lanczos3 optimized. It will go away with bilinear. Btw. You can also try: "VAAPI-BOB" and report how that worked out.

[tavoc]

If I remember right from my test yesterday Kodi crashes or goes back to the desktop enviroment if I use vaapi-motion compensate (MCDI and MADI) (I think because you stated "VAAPI-MCDI is not available on a Baytrail as this is an IVB based chip"...). Will try vaapi-BOB tonight and compare the image-quality.

Are skipped frames a bad thing? I can not complain at the image quality at the moment. It is much better with the 3.18 Kernel and the wsnipex ppa than 3.13 Kernel shipped with Xbmcbuntu. (channel Logos are not jumping any more and I do not need a slight overscan).

[fritsch]

Yeah, skiped frames are a bad thing, cause not all the frames we want to display get actually rendered.

You need the wsnipex vaapi ppa - And for baytrail I think I have to suggest Kernel: 3.16.7 from Ubuntu Mainline, cause of an upstream regression.

[tavoc]

This would be this one?

kernel.ubuntu.com/~kernel-ppa/mainline/v3.16.7-utopic/linux-headers-3.16.7-031607_3.16.7-031607.201410301735_all.deb \

kernel.ubuntu.com/~kernel-ppa/mainline/v3.16.7-utopic/linux-headers-3.16.7-031607-generic_3.16.7-031607.201410301735_amd64.deb \

kernel.ubuntu.com/~kernel-ppa/mainline/v3.16.7-utopic/linux-image-3.16.7-031607-generic_3.16.7-031607.201410301735_amd64.deb


I try to do a few tests this weekend with the suggested methods and kernels (starting with 3.18)

[fritsch]

No Tests needed. BYT has stability issues with 3.17+

[laric]

Suspend and Resume is optional - whenever the AVR is not yet up, you will get problems, a custom xorg.conf or a custom EDID will help.
I updated the kernel to 3.18.1 see first post.

How do I use an edid file with a intel haswell chip?

[fritsch]

google: drm_kms_helper intel

[tavoc]

Hi fritsch,

I have done some tests anyway, seems you were right. Baytrail can only cope with Deinterlace and Bilinear. Every other advanced Scaling method results in skipped frames. It does not depend on the deinterlace method, there I can choose Vappi-bob, automatic, deinterlace, none without problems.

Strange thing is that with gotham and without the modifications lanczos2 was possible without skipped frames (if I remember correctly).

Regarding the kernel problem: I still use your 3.18, and as you predicted I got some hangups yesterday evening (every 10-15 minutes). The day before everything was fine.
Therefore I deactivated all C-States in BIOS. Seems to work.


Can I use the image improvements with the Kernel 3.16.7 and activated C-States?