Mailing List Archive

NVidia Pixellation
On 07/30/2018 04:26 PM, Aman Gupta wrote:
>
> Aman, do you have a good connection at Nvidia?  My local ABC station
> seems to have again started sending streams that show corruption when
> decoded with MediaCodec (and VDPAU, for that matter), but not when
> decoded with ffmpeg software.  I have a short sample that I can
> provide.
>
>
> Yes I can forward the bug report and sample if you email them to me.
> Please test the sample with the latest Exoplayer and make sure it
> reproduces there so they can investigate easily.
>
>
I have a channel that has this problem. The local NBC station changed
hands at the beginning of 2017 and since then the stream gets periodic
pixellation but only on NVidia. The pixellations are at exactly the same
places on Shield as with VDPAU on my low-end, $20 NVidia cards.
Apparently NVidia are using the same (faulty) software in the Shield as
they have used for years in the low-end GEForce 210.

These recordings play perfectly with software decoding, VAAPI, and
*_Raspberry Pi OpenMAX_*. ($30 Raspberry Pi works better than Shield?!?).

If you need a sample I can also provide one.

Peter
Re: NVidia Pixellation [ In reply to ]
On Mon, Jul 30, 2018 at 9:42 PM Peter Bennett <pb.mythtv@gmail.com> wrote:

> If you need a sample I can also provide one.

My recommendation would be to send all the
available samples with as much detail in the
bug report possible for each unique station,
in the case the causes are different (even if
the symptoms are the same).

There are just too many ways to generate
visual artifacts in the codec implementations.

FWIW, I would not be surprised if the same
encoder is being used at multiple stations
(there are only a handful of vendors that
broadcast engineers trust/specify), but there
are dozens of knobs to adjust, and it may
take a special person to adjust the knobs
just so.
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Mon, Jul 30, 2018 at 05:41:00PM -0400, Peter Bennett wrote:
> On 07/30/2018 04:26 PM, Aman Gupta wrote:
> >
> > Aman, do you have a good connection at Nvidia?? My local ABC station
> > seems to have again started sending streams that show corruption when
> > decoded with MediaCodec (and VDPAU, for that matter), but not when
> > decoded with ffmpeg software.? I have a short sample that I can
> > provide.
> >
> >
> > Yes I can forward the bug report and sample if you email them to me.
> > Please test the sample with the latest Exoplayer and make sure it
> > reproduces there so they can investigate easily.
> >
> >
> I have a channel that has this problem. The local NBC station changed hands
> at the beginning of 2017 and since then the stream gets periodic pixellation
> but only on NVidia. The pixellations are at exactly the same places on
> Shield as with VDPAU on my low-end, $20 NVidia cards. Apparently NVidia are
> using the same (faulty) software in the Shield as they have used for years
> in the low-end GEForce 210.
>
> These recordings play perfectly with software decoding, VAAPI, and
> *_Raspberry Pi OpenMAX_*. ($30 Raspberry Pi works better than Shield?!?).

Interestingly, when I first encountered this problem (probably 8 or
more years ago) older Nvidia cards didn't exhibit the problem. For
example, my GT 430 and and GT 730 showed the problem, but the previous
generation 9500 GT I had my mom using didn't show it. I reported the
problem back then and sent Nvidia several samples as requested, but it
never got fixed.

David
--
David Engel
david@istwok.net
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Mon, Jul 30, 2018 at 10:56 PM David Engel <david@istwok.net> wrote:
...
> Interestingly, when I first encountered this problem (probably 8 or
> more years ago) older Nvidia cards didn't exhibit the problem.

This is not entirely surprising.

Different VPU firmware in the various generations
of NVidia hardware operate differently. Depending
on the details the decode was done in the dedicated
"PureVideo" hardware, or partially offloaded into
the general purpose CUDA implementations on the
GPU, and some parts on the CPU itself. The VPU
hardware has undergone not only enhancements
(additional codec support, different levels and/or
profiles), but reportedly (by those with much better
testing capabilities than I possess) more subtle bug
fixes (for good/bad/ugly reasons, NVidia does not
document the bugs in older hardware variants).
Last I checked, PureVideo was up to feature set I
(although I don't know of anyone using a Titan V
GPU (at $3000) in their HTPC).
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Tue, Jul 31, 2018 at 01:35:09AM +0000, Gary Buhrmaster wrote:
> On Mon, Jul 30, 2018 at 10:56 PM David Engel <david@istwok.net> wrote:
> ...
> > Interestingly, when I first encountered this problem (probably 8 or
> > more years ago) older Nvidia cards didn't exhibit the problem.
>
> This is not entirely surprising.

I know. I also know this has all been discussed here before at least
once and probably more over the years.

David

> Different VPU firmware in the various generations
> of NVidia hardware operate differently. Depending
> on the details the decode was done in the dedicated
> "PureVideo" hardware, or partially offloaded into
> the general purpose CUDA implementations on the
> GPU, and some parts on the CPU itself. The VPU
> hardware has undergone not only enhancements
> (additional codec support, different levels and/or
> profiles), but reportedly (by those with much better
> testing capabilities than I possess) more subtle bug
> fixes (for good/bad/ugly reasons, NVidia does not
> document the bugs in older hardware variants).
> Last I checked, PureVideo was up to feature set I
> (although I don't know of anyone using a Titan V
> GPU (at $3000) in their HTPC).
> _______________________________________________
> mythtv-dev mailing list
> mythtv-dev@mythtv.org
> http://lists.mythtv.org/mailman/listinfo/mythtv-dev
> http://wiki.mythtv.org/Mailing_List_etiquette
> MythTV Forums: https://forum.mythtv.org

--
David Engel
david@istwok.net
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Tue, Jul 31, 2018 at 1:49 AM David Engel <david@istwok.net> wrote:

> I know. I also know this has all been discussed here before at least
> once and probably more over the years.

I tend to post (approximately) the same information
every time someone references the differences
between their nvidia cards and who did not
reference those previous posts. And that is mostly
because I am also too lazy to go back and check
to see if the people who posted their newest
observation were on the original responses, and
are therefore fully responsible for remembering it.
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Mon, Jul 30, 2018 at 2:41 PM Peter Bennett <pb.mythtv@gmail.com> wrote:

>
>
> On 07/30/2018 04:26 PM, Aman Gupta wrote:
>
> Aman, do you have a good connection at Nvidia? My local ABC station
>> seems to have again started sending streams that show corruption when
>> decoded with MediaCodec (and VDPAU, for that matter), but not when
>> decoded with ffmpeg software. I have a short sample that I can
>> provide.
>>
>
> Yes I can forward the bug report and sample if you email them to me.
> Please test the sample with the latest Exoplayer and make sure it
> reproduces there so they can investigate easily.
>
>
>>
>> I have a channel that has this problem. The local NBC station changed
> hands at the beginning of 2017 and since then the stream gets periodic
> pixellation but only on NVidia. The pixellations are at exactly the same
> places on Shield as with VDPAU on my low-end, $20 NVidia cards. Apparently
> NVidia are using the same (faulty) software in the Shield as they have used
> for years in the low-end GEForce 210.
>

I hear back about my bug report. I forgot I had submitted a similar report
several months ago as well. Apparently it's a known issue in the hardware
and there's no software fix possible. Pretty disappointing. =(

Aman


>
> These recordings play perfectly with software decoding, VAAPI, and *Raspberry
> Pi OpenMAX*. ($30 Raspberry Pi works better than Shield?!?).
>
> If you need a sample I can also provide one.
>
> Peter
>
Re: NVidia Pixellation [ In reply to ]
On Sun, Aug 05, 2018 at 11:20:56AM -0700, Aman Gupta wrote:
> On Mon, Jul 30, 2018 at 2:41 PM Peter Bennett <pb.mythtv@gmail.com> wrote:
>
> >
> >
> > On 07/30/2018 04:26 PM, Aman Gupta wrote:
> >
> > Aman, do you have a good connection at Nvidia? My local ABC station
> >> seems to have again started sending streams that show corruption when
> >> decoded with MediaCodec (and VDPAU, for that matter), but not when
> >> decoded with ffmpeg software. I have a short sample that I can
> >> provide.
> >>
> >
> > Yes I can forward the bug report and sample if you email them to me.
> > Please test the sample with the latest Exoplayer and make sure it
> > reproduces there so they can investigate easily.
> >
> >
> >>
> >> I have a channel that has this problem. The local NBC station changed
> > hands at the beginning of 2017 and since then the stream gets periodic
> > pixellation but only on NVidia. The pixellations are at exactly the same
> > places on Shield as with VDPAU on my low-end, $20 NVidia cards. Apparently
> > NVidia are using the same (faulty) software in the Shield as they have used
> > for years in the low-end GEForce 210.
> >
>
> I hear back about my bug report. I forgot I had submitted a similar report
> several months ago as well. Apparently it's a known issue in the hardware
> and there's no software fix possible. Pretty disappointing. =(

Yes, that's definitely disappointing. Can they say what conditions
cause the problem? I know the chances of anything changing are slim,
but perhaps we could pass the information on to our local TV stations.
I know I've seen the problem come and go in the past so it's something
they can control in their encoders.

David

>
> Aman
>
>
> >
> > These recordings play perfectly with software decoding, VAAPI, and *Raspberry
> > Pi OpenMAX*. ($30 Raspberry Pi works better than Shield?!?).
> >
> > If you need a sample I can also provide one.
> >
> > Peter
> >

--
David Engel
david@istwok.net
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Sun, Aug 05, 2018 at 02:41:03PM -0500, David Engel wrote:
> On Sun, Aug 05, 2018 at 11:20:56AM -0700, Aman Gupta wrote:
> > On Mon, Jul 30, 2018 at 2:41 PM Peter Bennett <pb.mythtv@gmail.com> wrote:
> > > On 07/30/2018 04:26 PM, Aman Gupta wrote:
> > >
> > > Aman, do you have a good connection at Nvidia? My local ABC station
> > >> seems to have again started sending streams that show corruption when
> > >> decoded with MediaCodec (and VDPAU, for that matter), but not when
> > >> decoded with ffmpeg software. I have a short sample that I can
> > >> provide.
> > >>
> > >
> > > Yes I can forward the bug report and sample if you email them to me.
> > > Please test the sample with the latest Exoplayer and make sure it
> > > reproduces there so they can investigate easily.
> > >
> > >
> > >>
> > >> I have a channel that has this problem. The local NBC station changed
> > > hands at the beginning of 2017 and since then the stream gets periodic
> > > pixellation but only on NVidia. The pixellations are at exactly the same
> > > places on Shield as with VDPAU on my low-end, $20 NVidia cards. Apparently
> > > NVidia are using the same (faulty) software in the Shield as they have used
> > > for years in the low-end GEForce 210.
> > >
> >
> > I hear back about my bug report. I forgot I had submitted a similar report
> > several months ago as well. Apparently it's a known issue in the hardware
> > and there's no software fix possible. Pretty disappointing. =(
>
> Yes, that's definitely disappointing. Can they say what conditions
> cause the problem? I know the chances of anything changing are slim,
> but perhaps we could pass the information on to our local TV stations.
> I know I've seen the problem come and go in the past so it's something
> they can control in their encoders.

Aman,

Have you had the chance to follow up on this? I'd really like to
report it to my local station.

David
--
David Engel
david@istwok.net
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Fri, Aug 31, 2018 at 8:33 AM David Engel <david@istwok.net> wrote:

> On Sun, Aug 05, 2018 at 02:41:03PM -0500, David Engel wrote:
> > On Sun, Aug 05, 2018 at 11:20:56AM -0700, Aman Gupta wrote:
> > > On Mon, Jul 30, 2018 at 2:41 PM Peter Bennett <pb.mythtv@gmail.com>
> wrote:
> > > > On 07/30/2018 04:26 PM, Aman Gupta wrote:
> > > >
> > > > Aman, do you have a good connection at Nvidia? My local ABC station
> > > >> seems to have again started sending streams that show corruption
> when
> > > >> decoded with MediaCodec (and VDPAU, for that matter), but not when
> > > >> decoded with ffmpeg software. I have a short sample that I can
> > > >> provide.
> > > >>
> > > >
> > > > Yes I can forward the bug report and sample if you email them to me.
> > > > Please test the sample with the latest Exoplayer and make sure it
> > > > reproduces there so they can investigate easily.
> > > >
> > > >
> > > >>
> > > >> I have a channel that has this problem. The local NBC station
> changed
> > > > hands at the beginning of 2017 and since then the stream gets
> periodic
> > > > pixellation but only on NVidia. The pixellations are at exactly the
> same
> > > > places on Shield as with VDPAU on my low-end, $20 NVidia cards.
> Apparently
> > > > NVidia are using the same (faulty) software in the Shield as they
> have used
> > > > for years in the low-end GEForce 210.
> > > >
> > >
> > > I hear back about my bug report. I forgot I had submitted a similar
> report
> > > several months ago as well. Apparently it's a known issue in the
> hardware
> > > and there's no software fix possible. Pretty disappointing. =(
> >
> > Yes, that's definitely disappointing. Can they say what conditions
> > cause the problem? I know the chances of anything changing are slim,
> > but perhaps we could pass the information on to our local TV stations.
> > I know I've seen the problem come and go in the past so it's something
> > they can control in their encoders.
>
> Aman,
>
> Have you had the chance to follow up on this? I'd really like to
> report it to my local station.


I didn't get a clear answer: "has to do with motion vectors going outside
of some normal limit"


>
> David
> --
> David Engel
> david@istwok.net
>
Re: NVidia Pixellation [ In reply to ]
On 08/31/2018 01:43 PM, Aman Gupta wrote:
> I didn't get a clear answer: "has to do with motion vectors going
> outside of some normal limit"
>
I imagine that if the transmitter changed it, that may result in a
reduction in quality, especially since they are working with less
bandwidth these days. That would mean reducing the quality for everybody
to satisfy a few NVidia users.

Peter
Re: NVidia Pixellation [ In reply to ]
On Fri, Aug 31, 2018 at 5:54 PM Peter Bennett <pb.mythtv@gmail.com> wrote:
>
>
>
> On 08/31/2018 01:43 PM, Aman Gupta wrote:
>
> I didn't get a clear answer: "has to do with motion vectors going outside of some normal limit"
>
> I imagine that if the transmitter changed it, that may result in a reduction in quality, especially since they are working with less bandwidth these days. That would mean reducing the quality for everybody to satisfy a few NVidia users.

Actually, they way I read it, there are a couple of possible
interpretations. The first is that the encoding is configured
out of spec (that would be a station encoder setting or
firmware issue, which may be able to be adjusted/corrected),
and he other is that the nvidia VPU is not capable of meeting
the spec (that would be an nvidia issue to fix, if it can be
fixed at all).

Unfortunately, the answer does not make it at all clear,
or what next steps are appropriate (other than trying to
get clarity regarding the intent of the response regarding
standards compliance).
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Fri, Aug 31, 2018 at 06:34:34PM +0000, Gary Buhrmaster wrote:
> On Fri, Aug 31, 2018 at 5:54 PM Peter Bennett <pb.mythtv@gmail.com> wrote:
> >
> >
> >
> > On 08/31/2018 01:43 PM, Aman Gupta wrote:
> >
> > I didn't get a clear answer: "has to do with motion vectors going outside of some normal limit"

Thanks. I'll guess I'll have to go with that.

> > I imagine that if the transmitter changed it, that may result in a reduction in quality, especially since they are working with less bandwidth these days. That would mean reducing the quality for everybody to satisfy a few NVidia users.
>
> Actually, they way I read it, there are a couple of possible
> interpretations. The first is that the encoding is configured
> out of spec (that would be a station encoder setting or
> firmware issue, which may be able to be adjusted/corrected),
> and he other is that the nvidia VPU is not capable of meeting
> the spec (that would be an nvidia issue to fix, if it can be
> fixed at all).

Right. If it's the former, I would have liked to be able to tell
specifically what they're doing wrong.

David

> Unfortunately, the answer does not make it at all clear,
> or what next steps are appropriate (other than trying to
> get clarity regarding the intent of the response regarding
> standards compliance).
> _______________________________________________
> mythtv-dev mailing list
> mythtv-dev@mythtv.org
> http://lists.mythtv.org/mailman/listinfo/mythtv-dev
> http://wiki.mythtv.org/Mailing_List_etiquette
> MythTV Forums: https://forum.mythtv.org

--
David Engel
david@istwok.net
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Fri, Aug 31, 2018 at 10:54 AM Peter Bennett <pb.mythtv@gmail.com> wrote:

>
>
> On 08/31/2018 01:43 PM, Aman Gupta wrote:
>
> I didn't get a clear answer: "has to do with motion vectors going outside
> of some normal limit"
>
> I imagine that if the transmitter changed it, that may result in a
> reduction in quality, especially since they are working with less bandwidth
> these days. That would mean reducing the quality for everybody to satisfy a
> few NVidia users.
>

Here's an acknowledgment of this bug on the NVIDIA forums by an employee:
https://forums.geforce.com/default/topic/1076537/shield-tv/pixelation-when-watching-certain-channel-s-/post/5945645/#5945645


>
> Peter
>
Re: NVidia Pixellation [ In reply to ]
On 12/29/18 12:30 PM, Aman Gupta wrote:
>
>
> On Fri, Aug 31, 2018 at 10:54 AM Peter Bennett <pb.mythtv@gmail.com
> <mailto:pb.mythtv@gmail.com>> wrote:
>
>
>
> On 08/31/2018 01:43 PM, Aman Gupta wrote:
>> I didn't get a clear answer: "has to do with motion vectors going
>> outside of some normal limit"
>>
> I imagine that if the transmitter changed it, that may result in a
> reduction in quality, especially since they are working with less
> bandwidth these days. That would mean reducing the quality for
> everybody to satisfy a few NVidia users.
>
>
> Here's an acknowledgment of this bug on the NVIDIA forums by an employee:
> https://forums.geforce.com/default/topic/1076537/shield-tv/pixelation-when-watching-certain-channel-s-/post/5945645/#5945645
>
>
>
> Peter
>

Thanks for the update. It is interesting that they somehow think it is a
silicondust problem, but anybody using mpeg2 would be affected. With
MythTV, we are able to use ffmpeg software decoding (if we set an
appropriate number of threads) without any adverse effect. I notice that
others report that kodi with software decoding is unacceptable on the
shield. They must not be using ffmpeg for software decoding, or maybe
are only using one thread.

Peter
Re: NVidia Pixellation [ In reply to ]
On Sat, Dec 29, 2018 at 12:53:14PM -0500, Peter Bennett wrote:
>
>
> On 12/29/18 12:30 PM, Aman Gupta wrote:
> >
> >
> > On Fri, Aug 31, 2018 at 10:54 AM Peter Bennett <pb.mythtv@gmail.com
> > <mailto:pb.mythtv@gmail.com>> wrote:
> >
> >
> >
> > On 08/31/2018 01:43 PM, Aman Gupta wrote:
> > > I didn't get a clear answer: "has to do with motion vectors going
> > > outside of some normal limit"
> > >
> > I imagine that if the transmitter changed it, that may result in a
> > reduction in quality, especially since they are working with less
> > bandwidth these days. That would mean reducing the quality for
> > everybody to satisfy a few NVidia users.
> >
> >
> > Here's an acknowledgment of this bug on the NVIDIA forums by an employee:
> > https://forums.geforce.com/default/topic/1076537/shield-tv/pixelation-when-watching-certain-channel-s-/post/5945645/#5945645
> >
> >
> >
> > Peter
> >
>
> Thanks for the update. It is interesting that they somehow think it is a
> silicondust problem, but anybody using mpeg2 would be affected. With MythTV,
> we are able to use ffmpeg software decoding (if we set an appropriate number
> of threads) without any adverse effect. I notice that others report that
> kodi with software decoding is unacceptable on the shield. They must not be
> using ffmpeg for software decoding, or maybe are only using one thread.

Yes, thanks much, Aman. I'll likely chime in on that thread when I
get time to point out that A) it's a very looooooooooooong, standing
issue, not just with the Shield, B) it's also not limited to just
Silicon Dust's app and C) we worked around it as Peter notes.

David
--
David Engel
david@istwok.net
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org
Re: NVidia Pixellation [ In reply to ]
On Sat, Dec 29, 2018 at 5:53 PM Peter Bennett <pb.mythtv@gmail.com> wrote:
> ... It is interesting that they somehow think it is a silicondust problem

I doubt that nVidia engineers believe it is only a
SiliconDust issue, but SD is (likely) one of the
few corporate contractual engagements using
MPEG-2, protected path, and MCDI, making it
more of an issue, and getting traction in the
org mgmt. And without hardware MCDI the
result is not good enough for people who care
about quality and receive interlaced content,
and can tell the difference; the only good news
is that a not insignificant number of people
either can't tell the difference, or have their TV's
set to muddy the picture so that it all looks
equally bad, or simply are in denial (cheap is
good).
_______________________________________________
mythtv-dev mailing list
mythtv-dev@mythtv.org
http://lists.mythtv.org/mailman/listinfo/mythtv-dev
http://wiki.mythtv.org/Mailing_List_etiquette
MythTV Forums: https://forum.mythtv.org