HDTV Encoding

[BlueIce]

Last night I finally installed my antenna, and recorded some shows fine, but when I go to encode it to XviD from the HD MPEG the ATI HDTV Wonder spits out, it encodes at 10 fps with both Gordian Knot and Auto Gordian Knot. My PC specs are: AMD Athlon XP 2600+, 2x 512 MB DDR PC 2100, ASUS A7N8X-LA motherboard, ATI HDTV Wonder, ATI Radeon 9700 Pro, all at stock speeds and all with the newest drivers and software. Is this average for my system? or is there another method I can use to encode to XviD.
The settings I was using were 624x352,(source res is 1920x1088) and 128 kbps vb3 mp3.
and here is the auto gk log incase it is of use

::edit:: the attachment using this forum didn't work, i think cause an admin has to approve it so either way i uploaded it to my site http://www.modyourshit.com/files/WPTV-HD%201_agk.txt

[Latexxx]

Resizing takes all your processing power. There is no way around it. If you want to try something, try without resizing (encode at 1920x1088) but it is likely to be even slower.

[Sulik]

You can use the ATI MMC Media Library to encode directly to DivX, or to a high-bitrate intermediate format before encoding to Xvid. I found that it is much faster this way.

[BlueIce]

thanks for the help :)
even though it didn't help :p

[Teegedeck]

If you resize to 624x352, anyway, you could probably take some strain off your cpu by using "ReduceBy2()" beforehand, which downsizes the 1920x1088 to 960x544 by omitting every second line, deinterlacing the picture cleanly as a sideeffect. The deinterlaced picture should look even better than with KernelDeint, which you currently seem to use and which requires lots of processing time, and artefact-free. Lanczos should also have an easier job then.

There might be need for IVTC, of course.

[Didée]

Originally posted by Teegedeck
take some strain off your cpu by using "ReduceBy2()" beforehand, which downsizes the 1920x1088 to 960x544 by omitting every second line
Correction. "ReduceBy2()" will not omitt any lines, but reduce the framesize by internally calling "BilinearResize()". In case of laced input, it will therefore produce blending.

For clean field cancelling, use "SeparateFields().SelectEven()".

To get it as fast as possible, one could try "Pointresize(960,544)" before doing the final resizing. It's a little crude, but fast.

[Teegedeck]

I stand corrected. Thanks, Didée.

[BlueIce]

you can probably tell by me using gordian knot and auto gordian knot that i am not very good at writing avisynth scripts, but didee you seem to, if it is not too much to ask for, what would be a good avs for me to use to encode the hd mpeg?

[unixfs]

a couple of weeks ago a very useful function was added to libavcodec
decoders kow resolution decoding.
When you enable this option the library uses modified dcts to decode directly at 1/2, 1/4 or 1/8 resolution, so
1) decoding is much faster (I can watch smoothly all hdtv samples I have without frame loss)
2) you don't need to resize
3) quality may suffer :(

Obviously you will still want to deinterlace...

How to use it:
mplayer -vc ffmpeg12 -lavdopts lowres=1 movie.ts
or
mencoder -vc ffmpeg12 -lavdopts lowres=1 -ovc xvid -xvidencopts <your options> -oac copy movie.ts