[mythtv-users] Hard drive failure...

JC johnmythtv at crombe.com
Thu Mar 6 17:01:51 UTC 2003


At 06:07 PM 3/4/2003 -0800, Mattias Larsson wrote:
[...]
>I'm quite happy with the 480x480 resolution the TiVo provides me with, so 
>I'm certain that will be enough. I figure specing the backend for a bit 
>more will give me some headroom so I wont have to suffer dropped frames. I 
>do have a number of reasonably fast computers to try as backends, for 
>example, I could probably use the 2200 XP and test it. What I don't have 
>is a midrange machine, I have slow machines (PII-400, PII-450 and such), 
>so I dont really know what will be required for the frontend.

I don't know enough about backend requirements (yet) to comment there, but 
I think you are being way too conservative about the specs for the 
"frontend" machine (I mean in your earlier message, where you talked about 
using a 1.7 GHz Celeron). The largest MPEG4 test file I have laying around 
is about 720x330 (a 16x9 DVD rip). Mounting it over a 100 Mbps NFS 
connection, and playing it using xine, I can display it quite smoothly, 
fullscreen, on a K6-300 running with 64 MB of RAM ... as long as I don't 
expect the system to do anything else demanding at the same time. Running 
"top" via an ssh session shows CPU utilization bouncing around in the 
70-75% range.

This system was literally built from the junkbox here, except for using 
modern video (Radeon) and sound (Soundblaster) cards. It was the video card 
that dictated my choice of test machine, BTW ... it's the slowest mobo/CPU 
I have with an AGP slot. Having a video card whose X driver supports 
efficient xVideo is key to getting away with this, I think.

The other key is keeping extraneous demands off the CPU. In doing your 
tests on Windows machines, you give in too much to the high overhead that 
Windows imposes on systems. Carefully tailored Linux systems can be much 
more efficient, and even sloppy ones are probably better, than stock 
Windows installs. So I second the suggestion that you do your tests at 
least using Linux systems and software, even if you don't go immediately to 
MythTV. Otherwise, you'll overspec badly.

Unfortunately, I don't have a ready way to measure power consumption, your 
other concern ... can you suggest something I might try on that score?





More information about the mythtv-users mailing list