[mythtv-users] Non-continuous picture when watching TV in MythTV

[Ray Olszewski]

At 02:19 PM 3/3/2003 -0500, Andy Davidoff wrote:
>Part of what makes SCSI disks so much faster in high load situations is
>that the data doesn't have to pass through the CPU as it does under IDE.
>Buying a faster CPU or a 7200rpm IDE disk won't alter this truth.

But turning dma on for IDE/UDMA hard disks will change this, won't it? It 
certainly caused major improvements in my capture capabilities.

>I'm often surprised at how much of my laptop's CPU is consumed by tasks
>like burning CDs.  It stands to reason that Myth imposes much higher load.
>
>If you can't afford big SCSI, I bet you'd still benefit from putting your
>LiveTV buffer on a small SCSI disk and using cheap IDE for recordings.
[old stuff deleted]

The other odd thing I'm seeing here in tests of various Linux 
video-recording apps is that raw CPU speed matters far more than CPU type. 
I'd expected, for example, that a 1 GHz P-III and a 1.7 GHz Celeron would 
perform about the same, but in fact the Celeron captures and encodes (in 
tests that use vcr, libavifile, and the win32 codecs - 320x240 NTSC, about 
500 MB/hour) at about half the CPU load of the P-III. I can even get, with 
this software, acceptable recordings from a Celeron 533 (though it runs at 
about 85% CPU load).

Perhaps there are some least-common-denominator issues with the 
pre-compiled apps (in Debian) that cause them not to take proper advantage 
of the better CPUs? How do the pre-compiled .debs for MythTV fare in this 
regard?