[mythtv-users] Will moving to DDR-400 memory from DDR-333 improve performance?

[Chris Pinkham]

* On Mon Oct 30, 2006 at 02:10:38PM -0500, Brian J. Murrell wrote:
> Of course not.  But do you really think the kernel and a few linux
> utilities are gobbling up the hundreds and hundreds of MB of ram that we
> are seeing?

Where are these hundreds of MB you are seeing?

Here's samples from my 2 backends and my main frontend.  Between all three
of them I'm barely using 165MB and that's 2 copies of mythbackend and one
copy of mythfrontend.  Command being run is:

	ps auxw | egrep "mythfront|mythback|RSS" | grep -v grep ; echo ; free

Master backend w/ 3 AverMedia M179 cards:

USER       PID %CPU %MEM   VSZ  RSS TTY      STAT START   TIME COMMAND
mythtv    2074  0.1 11.8 248100 30244 pts/2  S    Oct28   3:27 mythbackend

             total       used       free     shared    buffers     cached
Mem:        255208     245296       9912          0      31932     111540
-/+ buffers/cache:     101824     153384
Swap:       265032       2928     262104

OK, so we're using ~30MB there for mythbackend.

Now the slave backend with 2 air2pc ATSC HD cards:

USER       PID %CPU %MEM   VSZ  RSS TTY      STAT START   TIME COMMAND
mythtv   28147  1.4  4.0 206800 20920 pts/2  S    Oct28  42:24 mythbackend

             total       used       free     shared    buffers     cached
Mem:        516500     189156     327344          0      18868      63956
-/+ buffers/cache:     106332     410168
Swap:       265064          0     265064

OK, using a whopping 20MB there.

Now the main frontend:

USER       PID %CPU %MEM   VSZ  RSS TTY      STAT START   TIME COMMAND
mythtv    5717  3.6 22.7 242300 116720 pts/4 S    Oct08 1167:51 mythfrontend

             total       used       free     shared    buffers     cached
Mem:        512668     503240       9428          0      60868     169424
-/+ buffers/cache:     272948     239720
Swap:       265064      55420     209644

So, using about 116MB there for a grand total of 166MB on all three boxes.

Just because the VSZ of a program is 260MB doesn't mean it's using that, look
at the swap numbers in my 'free' command.  On the slave backend where VSZ is
200MB,  I don't have anything in swap at all, so just because VSZ is 200MB and
RSS is 20MB it doesn't mean that I'm going to have to swap in 180MB of data
or program the next time that something wants to use mythbackend.

> It just seems that 0.20 came with a much bigger footprint than 0.19 had.
> I just wonder why.  Perhaps there is no reasonable explanation and there
> is a bug somewhere.

Perhaps there are a lot of people out there jumping to conclusions about
how much memory Myth is actually using.

--
Chris