Dual-Core processor utilization?

Discussion in 'CloneDVD mobile' started by HTKatzmarek, Nov 21, 2007.

  1. HTKatzmarek

    HTKatzmarek Member

    I'm wondering if there's been any thought to making the mencoder process make use of multiple cores to speed up the process. This would reduce the time needed to encode, particularly on 3-pass operations.

    Thx for any comments
     
  2. sej7278

    sej7278 Well-Known Member

    slysoft don't write mencoder, and i think only a subset of the codecs used are multithreaded.

    i know x264 is multithreaded - and you can get an almost linear speedup, i.e. each core adds about 94% speed.

    xvid isn't multithreaded - well the bleeding-edge build is, but only uses about 1.5 cores, i.e. 2 cores is only about 50% faster and 4 cores makes no further difference.

    i don't think its possible to have core1 working on pass1 and core2 on pass2, it simply wouldn't make sense.
     
  3. Peer

    Peer Redfox Development Team Staff Member

    Yes, x264 is the only codec that really gains speed with multithreading in mencoder, but also the results didn't play on the PSP, so this multithreading approach was abandoned.

    MPEG4 speedup with multithread is less than 10%, actually most of the time 2 or 3% - most of the time one thread waits for the other thread to release some common resource.

    So it's useless for now..
     
  4. Peer

    Peer Redfox Development Team Staff Member

    The best way to speed up a 3pass encoding is to do a 2pass encoding :)

    Seriously: 3-pass is just not worth the effort. If you have a PC doing the job over night, it's fine, why not. But you'll still have trouble telling the difference from 2-pass.
     
  5. HTKatzmarek

    HTKatzmarek Member

    OK...thx for the input.
     
  6. Eltaco

    Eltaco Member

    I actually worked on ripping my entire DVD collection. I have a DVDROM and a DVDRW drive and would rip 2 DVD's at a time. I know that doesn't speed up a single encode. Using taskmanager I saw that durring a single encode it would rarely top above 50%, so while running 2 encodes would be slower total time than encoding individuals. Sorry if I am not making sense. It just was easier for me starting from scratch.