[mythtv-users] What do I have to do to get HD working?

John P Poet jppoet at gmail.com
Fri Apr 14 23:12:25 UTC 2006

On 4/14/06, Mark Lehrer <mark at knm.org> wrote:
> > I'm about at my wits end.  I just purchased a P4 3.2ghz HT to do HDTV
> > playback in software but no matter what I do, I get continual studdering
> > ...
> > happens on my MX4000 PCI video card system, my ATI 9100 IGP video card
> > system, and my Unichrome Pro video card system.
> >
> > Everything I read said that a P4 3.2 is enough to do HD but nothing I do
> > can get it to work.  I'm running the 2.6.15 smp kernel but that almost
> > seems to make it worse.  With the non-smp on my Unichrome system it
> Whoever said that a P4 3.2 is fast enough for HD doesn't know what
> they are talking about.  Get an AMD dual core.  I have a 3800 and it
> works well (It replaced a Prescott 3.2 that I used before after making
> the same mistake that you just made).
> However, with that being said, it should at least be fast enough to
> play back well at 30 frames/sec - if you use a deinterlacing method
> that discards every other field.

I have successfully used a P4 2.8GHz HT processor for HD without any
problems.  It is fast enough for either Kernel or BOB deinterlacing. 
For sound I was using an old SB Live! card.

If a HT 3.2GHz P4 was not fast enough, then you had something else wrong.

For instance, if you are using an nVidia video card, make sure that the

Option "NvAGP"

is set correclty.  If I had it set wrong for my motherboard, playback
stuttered badly.

Running a SMP kernel with the HyperThreaded P4 was a big win.  If I
tried running a non-SMP kernel, it did not fair as well.

Compiling Myth with --enable-proc-opt helps a great deal.

The 2.8GHz processor was *not* fast enough to flag commericals, record
and watch all at the same time.  It was also not fast enough to watch
at greater than 1.0x normal speed.

I have recently upgraded to a dual core so I can playback HD material
at greater than 1.0x normal speed.


More information about the mythtv-users mailing list