VideoHelp Forum




+ Reply to Thread
Results 1 to 3 of 3
  1. I've had my dual AMD rig for a few months but have recently started encoding video on it. Here are the relevant specs of my dual.

    2 x Athlon XP 1600+ @ 1667 MHz (unlocked and SMP modded)
    MSI K7D Master L (v1.5 BIOS)
    12.5 multiplier x 133 FSB
    1 stick 256mb Mushkin PC2400 DDR (non-ecc non-reg 2.5-3-3-6 timings)
    WD 200BB Hard Drive
    Windows 2000 Pro SP3

    I started encoding on the dual for (of course) one reason. TMPGenc is a mulththreaded app. So I've been tweaking with the performance a bit and have come upon something interesting and rather dissappointing regarding TMPGenc's multithread performance. When I got unexpected encoding times I decided to do some controlled testing. I've created an image to present my results. The details of the tests are explained below.



    The tests were all run with TMPGenc v2.59.47.155 Pro

    The video file was a 22m 23s divx avi
    640x480 resolution
    23.976 Frames per second
    The Audio was the original video's track extracted to a wav with virtualdub and encoded with TMPGenc's native encoder.

    The video was being encoded to 480x352 SVCD / 2 pass VBR / 1603 average bitrate.

    The difference between the simple encode and complex encode was filters. I ran no filters on the 'simple' encodes. For the 'complex' tests I turned on noise filtering (default, not high quality), sharpen edge, and simple color correction

    Okay now take another look at the results of my testing. When going from no multithreading to multithreading in the simple encode the encoding time is greatly cut down. Cpu utilization is acceptable and was generally hanging around 80% compared to 50% in the no mulththread test. With the complex encode however it is a much different story. The encode time is only cut by a small percentage and the cpu utilization stayed mostly at 50% with occasional jumps up to 70 - 80%. The complex multithreading test was certainly different from the no multithread test, but nothing near the speed increase I want from a multithreaded app.

    Here is the most interesting part

    Difference between multithread and no multithread tests

    Simple = 11m 42s
    Complex = 13m 53s

    Nearly the same aren't they? This says only one thing to me. TMPGeng is certainly a multithreaded app however only its core encoding processes actually use multithreads. Other encoding options are left to run as single threads just like they would have on a single CPU system.

    Here is what I want to know from you guys.

    What are your experiences with encoding on your dual CPU systems? Do other encoders (I'm primarily interested in AVI to SVCD) take better advantage of multithreading when encoding? What I truly want is an application that will encode my videos, filter them (TMPGenc has great filtering options and the equivalent is a MUST for any encoder I use) and take full (or better) advantage of both CPUs.

    Thanks for reading
    Quote Quote  
  2. Член BJ_M's Avatar
    Join Date
    Jul 2002
    Location
    Canada
    Search Comp PM
    most filters are not MP enabled on almost any app -- even very high end ones .. though everything else can be in the app ..
    Quote Quote  
  3. I don't use TMPGenc so this may not apply.....

    - Try to frame serve using avisynth (and/or) virtual dub. You could spread the filters out across all of the apps (if you use multiple filters). This should really crank up the cpu usage. If you use one 'big' filter (high settings), try to use a couple different ones across the apps.

    - Do the prefiltering in a vdub run as above, and save to mjpeg. Then encode. This may increase speed due to reduce memory usage, vs the bullet above.

    I do this with avisynth and vdub, on a dual 1900. I can definately get higher fps and cpu usage this way. 18fps in vdub and ~90% on both cpus for heavy filtering of captured cartoons. My dual P3 933 could only do about 7fps on the same filters. While I'm blathering about my personal experience, I thought I'd mention that I've been using the MainConcept encoder. BJ_M posted lots of good stuff about it. It produces quite a good encode (at default DVD VBR settings). It gets to ~90% on both cpus AND encodes at 2x real time on my machine. Yes, that would be a 1hr 40min movie in about 50min.

    I don't know why I ever wasted my time on cce 3 pass vbr. :P
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!