[MPlayer-users] Performance data points for embedded systems

[Felix von Leitner]

I am in the process of building a VIA C3 based mplayer appliance for my
living room.  The main goal is to have a silent (i.a. not "little" noise
-- _no_ noise!) box that can play movies over network or from a CD-R (or
DVD).  Of course, there are no silent CD or DVD drives, but the system
itself should be absolutely silent.

My previous experience with VIA C3 based systems shows that mplayer is
able to play a DivX 5 file from CD-R using the frame buffer device (i.e.
no xv!) on a 900 MHz C3 with top(1) showing 40% CPU idle and the passive
cooling being only hand warm.  However, top is not a good measurement
tool and that box had a power supply unit with a fan that caused some
air flow through the passive cooler.  Now I am planning to build a
fan-free box; I found a case with external notebook style PSU on
www.mini-itx.com.  My last experiment used a Flex-ATX mainboard, this
time I want to Mini-ITX one.

The problem is that the VIA EPIA mainboards (my current target platform)
can be had with passive cooling for up to 667 MHz and with active
cooling up to 933 MHz currently.  I am considering buying the 933 one
and removing the fan, but it would be better for my warranty if I bought
a 667 one and that also worked.

My question is: has anyone ever done something like this?  I found
several web pages showing that Windoze is too inefficient to play DivX
on the 667 one, I even found a web page that showed that the 900 MHz one
was too slow on Windows.  So the Windows data points are clearly
worthless since mplayer is much more efficient.

I ordered a small case and an EPIA board with 933 MHz C3 last week, but
it is apparently not available yet.  The board is by the way the new
generation EPIA board with DDR RAM (my old measurement did not even have
DDR RAM and you have to consider that the system even used an unified
memory framebuffer, so even less memory bandwidth is available to the
CPU).

Felix