[mythtv-users] Recommendations needed for new FE box

[David Engel]

On Fri, Jul 27, 2018 at 02:37:21PM -0500, Ashu Desai wrote:
> On Fri, Jul 27, 2018 at 2:04 PM Gary Buhrmaster <gary.buhrmaster at gmail.com>
> wrote:
> 
> > On Fri, Jul 27, 2018 at 5:48 PM Ashu Desai <ashu.desai at gmail.com> wrote:
> >
> > > I thought VDPAU was the "better" one here,
> >
> > For those of discerning tastes (i.e. the quality addicts),
> > in the de-interlacing world, what you want is MCDI(*)
> > (motion compensated de-interlacing).  MADI (motion
> > adaptive de-interlacing) is a lower complexity
> > implementation which some find adequate.  With
> > the exception of extremely powerful systems, those
> > require hardware assistance(**).  There are also various
> > software de-interlacers, at varying levels of quality
> > (and software demands).
> >
> > While today all the major vendors have hardware
> > support of MCDI, that was not always true.  Nvidia
> > was, for years, considered the gold standard in the
> > HTPC world for video quality.  MythTV's implementation
> > of VDPAU included the MCDI support.
> >
> > Early gen Intel VPUs did not support advanced hardware
> > de-interlacing, and the original VAAPI also did not.  Later
> > Intel hardware added (first) MADI, and eventually MCDI.
> > A later version of VAAPI now supports those, but it uses
> > a different architectural approach that MythTV was never
> > updated to use (a couple of people have reportedly
> > looked at it, but it was determined to require enough
> > work (i.e. time) that it was simply not done).
> >
> > Over the years FFMpeg has add support for advanced
> > hardware accelerated decode and de-interlacing,
> > including for Intel based MCDI (they also have a lot more
> > devs dedicated to the video pipelines they support).
> > However, as David Engel points out, MythTV did not
> > utilize FFMpeg in ways that took advantage of some
> > of those new/improved capabilities, and that the work
> > by Peter Bennett may(***) be able to be leveraged to
> > support in the future.
> >
> >
> >
> >
> >
> > (*) There are subtle differences between hardware
> > implementations of MCDI, but while it is possible to
> > observe differences in synthetics, it is the rare
> > individual that will be able to see the difference in
> > common scenarios in move vendors latest hardware.
> > On the other hand, many can see the difference
> > between MCDI and blend (a simple software
> > de-interlacer).
> >
> > (**) And hardware assistance to run in any reasonable
> > power budget.
> >
> > (***) "May" as in someone still has to do the work.
> > Typically that happens when someone with the
> > appropriate skills has the itch to do the work.
> >
> >
> Thank you!
> 
> My problem is I don't know the difference say between Raspberry Pi 3 B+
> video output vs say NUC with Pentium J5005 - i see multiple NUC on amazon
> with under $200
> 
> So just trying to see which way to go.

For what do you want to use it?  MythTV, timestretch, YouTube, Prime
Video, Playstation Vue, other streaming service, other HTPC function?
Some boxes can do some of these, but no single box can really do them
all well.  In other words, it all depends on what *you* want it to do.

David
-- 
David Engel
david at istwok.net