[mythtv-users] Time to take the next step...

[Yeechang Lee]

VCRAddict <MythTV_01 at appropriate-tech.net> says:
> We were talking about several different things, and those two comments were
> in response to two different statements made by Allen Edwards.
> Specifically, the first was in response to his contention that an AMD
> Athlon XP 3000+ would be inadequate for my proposed Back-End-Only system
> (which I find implausible, to say the least);

His message was somewhat ambiguous, but I disagree on the proper
reading. From
(<URL:http://www.gossamer-threads.com/lists/mythtv/users/357580#357580>):

> I think you need more processor power for ATSC system. I have a
> frontend only with a 2.4GHz P4 and it is not quite fast enough even
> running XvMC (which you do not want to run). My FE/BE runs a 5400+
> and it is fine.

Given how he talks about how his 2.4GHz Pentium 4 frontend is
inadequate "for ATSC" (which in the US is more or less syononymous
with "high definition"), he was (as I read it) referring to your
expectation that a 1.6GHz Pentium 4 would be sufficient as a HD
frontend. Allen, can you clear this up for us?

> As for whether or not the existing FE will prove adequate for playback of
> the stuff the new Back-End captures... that remains to be seen; but I
> suspect that I'll *have* to use XvMC for it to work acceptably.

Again, your 1.6GHz Pentium 4 frontend just isn't going to work today
for native HD playback, with or without XvMC.

> Even with XvMC?  I find that surprising, given other comments & reports
> I've read.  I seem to recall a lot of folks using things like those little
> EPIA micro boards for super-quiet dedicated FE-only systems.

Not with HD playback.

That said, I admit to not being personally familiar with the current
state of the VIA Unichrome drivers; last I heard they were pretty darn
good, but I don't know how good they are with HD, so I could very well
be wrong here.

> ISTM that if a VIA C3 can work even moderately well, 1.6GHz P4
> should not have a problem.  This chart:
> 
>     <http://www.mythtv.org/wiki/index.php/XvMC:Example_CPU_Savings>
> 
> also seems to imply that I'll have enough raw hardware horsepower, if
> perhaps little or none to spare.

I must disagree. No VIA users report using HD. No Intel users report
doing satisfactory HD playback with or without XvMC using anything as
slow as your Pentium 4 1.6GHz. (I freely admit I've never learned how
to read AMD performance figures relative to Intel so won't comment
there, but the AMD examples look more promising to this untrained
eye.)

>  > > That is not the situation I'm dealing with.  The new system
>  > > will be a Back-End ONLY.  My understanding is that this reduces
>  > > the host system CPU/memory requirements *drastically*, as
>  > > compared to a FE/BE system.
>  > 
>  > Memory, yes.
>  > 
>  > CPU? Not so much. As noted, recording ATSC streams from a HDHomerun is
>  > very easy from a processing perspective.
> 
> Now it seems to be you who is contradicting himself.
> 
> As noted several times so far in this thread, "recording ATSC streams" is
> *all* this box is destined to do.

No contradiction. Same misunderstanding as earlier, in which Allen was
(as was I) referring to your 1.6GHz current frontend/backend and I
followed suit. I took "host system" in your first quote above to mean
the same 1.6GHz system, and for your overall statement to mean that
you believed that offloading the backend functionality from it would
provide enough horsepower to play HD. We were speaking about apples
and oranges, it turns out.

To clarify, my point was that removing the backend functionality from
the 1.6GHz box would help relieve memory usage more (much more,
depending on how big your database gets) than CPU usage (it'll help
some, not a lot).

-- 
Frontend:		P4 3.0GHz, 1.5TB software RAID 5 array
Backend:		Quad-core Xeon 1.6GHz, 6.6TB sw RAID 6
Video inputs:		Four high-definition over FireWire/OTA
Accessories:		47" 1080p LCD, 5.1 digital, and MX-600