0 Replies Latest reply on Jun 24, 2013 2:21 PM by MD Optofonik

    Low budget system test with “unlocked” video card and proxy files.

    MD Optofonik Level 1

      In an effort to understand the shortcomings of using Premiere CS6 on my older computer I spent the weekend learning about creating and using proxy files, how long it takes to transcode (using some old footage encoded to m2ts aka AVCHD-Lite), and how well my "unlocked" video cards play with MPE. I wanted to significantly reduce the file sizes, tried FLV first, than settled on mpeg2 with a quality setting of "1". Transcoding was much quicker with mpeg2 but with a lot of files it still toook awhile. Next I looked at the performance graphs in Task Manager to get a sense of how much the two different files' formats were taxing my computer during playback. I also alternated between MPE setting since my card is and "unlocked" card and not offically supported.

       

      One thing I noticed during the following is that when MPE “GPU Acceleration” is turned on, one core is loaded more than the others.

       

      When MPE is set to “Software Only”, all cores are loaded equally.

       

      As such it appears that the “Software Only” setting produces better results when looking at Task Manager's performance graphs. Just looking at the graphs there was about a 10% difference overall between using m2ts files with the “GPU Acceleration” setting vs mpg2 files with the “Software Only” setting. However, using “GPU Acceleration” appears to only use one core thereby allowing the others to be utilized elsewhere. I don't know much about multi-threaded core allocation so I wonder if that sort of thing even makes a difference. I have seen, in another program, evidence that the effect on a program's performance by such changes to core allocation can be significant. None of this is very scientific but it worked for me, I have certainly leaned a fair bit over the weekend; maybe someone else will find this useful.

       

      My conclusion is that transcoding original m2ts files to proxy files for editing, even on a low spec box like mine, probably isn't worth the time. Other formats are certainly another matter.

       

       

       

      Here are my notes:

       

      Using m2ts files at ¼ playback resolution with MPE “on” RAM is at 2.87GB and my CPU is running around 12% with dips to 8% and spikes to 15%. At “full” resolution my CPU is running around 16% with dips to 14% and spikes to 20%. RAM is at 2.93.

       

      With MPE “off” at ¼ playback resolution the RAM is still at 2.87GB and CPU fluctuates a lot more but mostly around 10% with dips as low as 8% and the same spikes at 15%. However, at ¼ resolution the preview suffers pronounced jaggies. Switching to “Full” resolution fixes most of the jaggies and increases RAM to 3.08GB; CPU hovers around 12% with dips to 10% and spike up to 17%. So, the MPE doesn't do much with the numbers but the preview pane looks better at lower resolutions.

       

      With my MPEG2 proxy files at ¼ resolution and MPE “on” Ram is at 2.85 and my CPU is running around10% with dips to 6% and spikes to 13%. At full resolution my CPU is running around 12% with dips to 8% and spike to 16% RAM is at 3.03.

       

      With MPE “off” at ¼ resolution RAM is at 2.87GB and my CPU is running around  8% with dips to 6% and spike to 11%. At full resolution my CPU is running around 9% with dips to 6% and spikes to 13% RAM is at 3.03 with no jaggies.

       

       

       

      Original m2ts files:

       

      M2ts MPE on ¼ 2.87 RAM 12-15% CPU

       

      M2ts MPE on full 2.93 RAM 16-20% CPU

       

       

      M2ts MPE off ¼ 2.87 RAM 10-15% CPU (jaggies)

       

      M2ts MPE off full 3.08  RAM 12-17% CPU

       

       

       

       

      Proxy MPEG2 files (set to “1” quality in ME):

       

       

      Mpg MPE on ¼ 2.85 RAM 10-13% CPU

       

      Mpg MPE on full 2.85 RAM 12-16% CPU

       

       

      Mpg MPE off ¼ 2.87 RAM 6-11% CPU

       

      Mpg MPE off full 2.85 RAM 9-13% CPU