Log in

View Full Version : What's Your Longest Encode?


EpheMeroN
24th September 2005, 18:51
I just finished probably the longest encode I have ever done on a video. It ran the entire time at about 0.5fps and was just slooooooooooooow. I wondered if anyone else here has had some harsh experiences like mine! Mine took over 110 hours. It was done on a DeLL Inspiron 1250 Laptop. It uses a P4M @ 1.7ghz, 256mb pc2100 ddr ram, and a 30gb hdd @ 5400rpm. I did a standard 2-pass encode using the BEST profile.

http://img279.imageshack.us/img279/7074/untitled8lo.jpg

Kostarum Rex Persia
24th September 2005, 19:14
Ok,but you missed to tell us which machine you have,and how much RAM?
Do you ever try to use Exaustive Search mode in x264 codec,in 2 or 3 pass?

mg262
25th September 2005, 01:10
If you are filtering heavily, you end up reapplying the filters on each pass, which is rather time-expensive. You can get round it by saving the output of your script (movie.avs above) to a lossless format, and then feeding that into the encoder.

Inventive Software
26th September 2005, 14:20
I'm doing one now that's projected to take around 10-11 hours. I started it early this morning, checked on it an hour ago and it wasn't even half-way there.

The longest I've tolerated is an XviD 2-pass of "Pirates Of The Caribbean". That took around 20 hours in total. Bearing in mind the rig I have, that's not bad (DivX is about 5% faster), but I'm not doing AVC encodes on it any time soon.

Having seen the picture, I can imagine why it took so long. MPEG-2 encoding on a Dell has never been brilliant, but also your processor is not particularly tuned to video encoding. Pentium-M processors are not brilliant for encoding video, let alone in MPEG-2.

Also, use QuEnc. Much faster! ;)

bill_baroud
26th September 2005, 16:01
Uhuh, in the good old Flaskmpeg days (wwoaaah :D), on my amazing Celeron 466Mhz, i ran an encode up to 40hours. I remember i stopped it in the middle of the 2nd pass because it was going slower and slower, Win98 wouldn't free ram correctly...

movax
26th September 2005, 18:43
Like mg262 said, if you're filtering heavily, make a lossless of it, and then you can run 2pass XviD on that in about a hour or so. (standard anime episode, 15, 16 hours for a lossless, than a hour for the XviD).

Didée
27th September 2005, 09:06
I remember a job where one-passing the 50min source to lossless took 150 hours or so ... no-one is impressed if the encoding time isn't measured in weeks. ;)

Inventive Software
27th September 2005, 11:18
Ooooh! That's gotta hurt!

mg262
27th September 2005, 11:20
Ltsmc?

Didée
27th September 2005, 12:49
No ... a concert recording with lots of issues. AR 1.85, non-anamorphly squeezed into 2/3-D1, shot on video but one field kicked out (-> 464x200 effective resolution, aliasing included), blurry, fuzzy, rather noisy. Text overlays put on "grey curtains" & crossing the black borders.
Figuring the script(s) took a multiple of the encoding time - and it should be done all over again, with lots of new means available since then...

mg262
27th September 2005, 13:21
Figuring the script(s) took a multiple of the encoding time - and it should be done all over again, with lots of new means available since then...Isn't that always true?! :) These days, I seem to have got round the problem by not encoding anything any more...

Soulhunter
27th September 2005, 14:08
Athlon XP 2600+ & 2800+ / 1024MB RAM / Just the usual mfps madness... ^^

http://img264.imageshack.us/img264/7566/doh19wf.png (http://imageshack.us)http://img264.imageshack.us/img264/8562/doh23wj.png (http://imageshack.us)

http://img264.imageshack.us/img264/9829/fft3ddenoising14lj.png (http://imageshack.us)http://img264.imageshack.us/img264/1093/fft3ddenoising24pg.png (http://imageshack.us)

http://img264.imageshack.us/img264/5210/mfps052tc.png (http://imageshack.us)http://img264.imageshack.us/img264/9034/mfps047tj.png (http://imageshack.us)

Bye

SirCanealot
27th September 2005, 19:05
but also your processor is not particularly tuned to video encoding. Pentium-M processors are not brilliant for encoding video, let alone in MPEG-2.



I thought Pentium Ms are pretty good. My 1.5ghz Pentium M has been pretty solid the few times I've used it for encoding. Was only a little slower than the 2.1ghz Atholn I had at the time. 'corse my 2.3ghz Atholn 64 is now a chunk faster than my Pentium M :D

Tuesday
27th September 2005, 20:10
I've done a few encodes using Didée's IIP, i think the source was Terminator III and to encode a temporary file in VBLE, encoding time was ~72 hours on my P4 3Ghz, 1024mb DDR400 machine. I think the temp file was well over 30gig also...

Inventive Software
29th September 2005, 09:22
The encode I mentioned earlier... that took 11:03:26. And that was only on the 1st Pass! Gotta do the second today... that's predicted about 10 hours.

Why is it that MPEG-2 encoders are so badly written sometimes? It's like some people just think: "OK, lets make an MPEG-2 encoder, but we'll do it THIS way instead" and never test their product to make sure the blooming thing works.

That said, I've used QuEnc many times and it's often out-performed the codecs I've used to backup the material.

OK, I take what I said about Pentium-Ms back, sorta. They're not brilliant but they do the job relatively well. It also depends what encoder you are using to encode. Like I said, some of them are written so badly they shouldn't be allowed to exist.

Mug Funky
29th September 2005, 10:04
Why is it that MPEG-2 encoders are so badly written sometimes? It's like some people just think: "OK, lets make an MPEG-2 encoder, but we'll do it THIS way instead

like TMPGenc... "let's make it work only on rgb24... it makes sense what with all sane formats being yuv..."

quenc is awesome :)

[edit]

btw, i'm doing some long encodes at the moment - 16mm transfers. while they are good sources compared to the alternatives available (often none), they are so grainy and flickery that i have to throw a slew of filters at it to repress the blocks. it's all worth it though. 2.4 fps...

Inventive Software
29th September 2005, 10:19
quenc is awesome :)

Agreed! :D

Inventive Software
3rd October 2005, 13:50
I just worked it out. Turns out your encode took 2.81 seconds per frame!!!! That just shouldn't be allowed!

danpos
3rd October 2005, 19:15
@All

Speaking about encoders out, have you given a try to new AQE beta (http://www.autoqmatenc.com/BetaTesters.html) by SAPSTAR ? I've done several tests with, in different scenarios and I've to say that results are great and it's very fast as well.

Cya!