performance improvements?

[aditya_lalla]

Disclaimer: i'm something of a newbie at encoding (one movie old ...) and am looking for a little detailed info. if you feel that any of the questions i have asked are already covered in other forum threads, please leave me a link. but i should also point out that i have seen the other topics being discussed, and am pretty sure none of my questions have been answered (atleast completely) in any of them ...

for my first rip, i used the gordian knot+meGUI as described here http://kitty.110mb.com/dvd2avc/, and am very very happy with the results ... but the whole process took me approximately 26 hrs. i used the nero AAC codec (audio) + h.264 (video) and put it all into an .mp4 container.

system specs:
amd athlon 2600+ (1.9 ghz)
512 mb ram ddr ram (166 mhz)
333 mhz fsb
ati radeon 9600 (256 mb ram) ... doubt this makes a difference though ....

1) what is the best way to improve encoding performance? does encoding performance scale in proportion to the number of cores (like rendering performance)? or does it depend purely on the horsepower available (cpu clock, fsb, ram etc.). does having multiple cores make any difference at all to performance? is encoding multi-threaded?

2) are there any specialised hardware or (hardware + software) solutions for speeding up the process? i've heard that avivo used with the radeon x1600 (or better) is supposed to help, but how much?

3) am i using the best software bundle available? or is the gordian knot + meGUI combination no better/worse than the other options out there? here i'm interested in just the speed and the final output quality.

4) is there a better codec/container which i should be using? again, purely in terms of speed and quality, with quality being far more important.

[buzzqw]

what is the best way to improve encoding performance?

i miss the profile used... since is the most important slowing factor

i've heard that avivo used with the radeon x1600 (or better) is supposed to help, but how much?

but in that case you will use the avivo codec, that is much worst then x264

the workflow could be much smaller, first of all after rip open megui and go in D2V creator and create audio and d2v, check both "on completition load files" and "and close", hit CLOSE

go to queue and click start , after some time the avisynth script creator pop up, click on autocrop, check "suggest resolution" , go on filter tab, click analyse , on noise filter select minimal noise, and check Color correction, check "On Save close and load to be encoded" and hit "Save" , choose the name fo avs

on input tab, select codec x264 and container MKV, , under video profile select CE-MainProfile , on audio ND AAC and NDAAC HE Multichannel 128, ckicl autoencode

select your storage medium and click queue
go on queue tab and click start :)

BHH

[aditya_lalla]

ah. i think you misunderstood the reason behind my query ... im not looking for a faster method, i want to know what kind of hardware would be best suited for converting my dvds to my desired format at the desired quality. ie., assuming all other factors and options staying constant, what hardware would give me the max fps. while ripping?

i suppose it was my last 2 questions that were misleading. i am very happy with the quality of the rip obtained. i'm just wondering if a better quality is possible with a different codec or if there's a software bundle out there that can do something of the same quality faster. i've tried out a few other options, and some of them were really sucky (even in the high quality mode ...)

[buzzqw]

for max fps ? look here http://forum.doom9.org/showthread.php?t=122318

the bost is faster/more cpu and faster ram

the software bundled don't make the difference, on your setup the preparation of audio and avs could take 1 hour but the encoding 25hour

and since is the codec x264 that is slow ... a different bundle with x264 cannot go faster (at equally x264 settings)

The varius all in one commercial encoder are mostly based on mencoder and if even true that goes faster is not true that look equal

BHH

[foxyshadis]

x264 scales nearly linearly with the # of cores/cpus, up to about 32. So the hardest part of the encoding is taken care of there. Then, avisynth needs a specially modified version (see the avisynth forum) to take advantage of multithreading; with that, you'll get a boost in the avisynth area as well.

In other words, lots of cpu cores makes the encoding go quite a bit faster. You can also cluster with one of the specialized x264 versions, but that depends on your network and other systems' performance.

[aditya_lalla]

nice! thanks. i hadn't realised they'd got the chart up. i suppose next time i won't go directly to the last page ... heh.

so i suppose that just leaves the question of how well encoding scales with the number of processors you throw at it ... and judging from the charts, overclocking the E6600 by 45% results in an almost proportionate 43% boost ... while the slightly higher clocked QX6700 give > 100% speed boost with 2x as many cores ....

hm. thats all my questions answered! thanks buzzqw :thanks:

ps: i don't suppose anyone would be able to tell me if future gpus would be able to encode? since they're already approaching 1 tflop (in sli/crossfire), can process a huge number of parallel threads and now that ati and nvidia are releasing CTM and CUDA frameworks ... it is possible.

[Blue_MiSfit]

Sure it's possible. We already have AviSynth filters (fft3dgpu) that run on the GPU and are blazingly fast (and also leave lots of CPU cycles for x264 :D)

Who knows, someday we might have x264_3dgpu :D