Test Report on x64 Builds

Discuss about generic usage of MediaCoder.

Moderator: HuggiL

Posts: 3116
Joined: Tue Jun 23, 2009 2:45 pm
Location: Sydney, Australia

Test Report on x64 Builds

Post by meRobs » Fri Nov 16, 2012 3:31 pm

Updated: 1 March 2014

All tests described below were conducted on the 64-bit versions of MediaCoder (Full Edition).
I had moved to a 64-bit Win7 machine and switched to the exclusive use of the x64 builds, from 5290. They are not only quicker, but, do not have the problem with Xvidcore that may occur in the 32-bit builds (viewtopic.php?f=28&t=12250).

Earlier tests on builds up to 5290, using 32-bit builds (x86), are described in: viewtopic.php?f=17&t=10383.
All builds from build 4045 to the most recent are available from: http://www.videohelp.com/tools/MediaCod ... s#download .

Also, I changed the Add Track tests to cover the bugs discovered in Dec 2012, which also apply to much earlier builds:
(1) missing frames in chapters (viewtopic.php?f=28&t=12368) and
(2) faulty PAR/DAR of the output (viewtopic.php?f=28&t=12366).

NOTE: as of build 5540, I ceased testing builds on a regular basis because the developer of MediaCoder seems more concerned with adding and working with new features, such as Segmental Video Encoding and H.265, rather than fixing 'existing' bugs, which of course is his right. In the meantime, I shall use build 5392, the last build that worked 'well' (Edit: still true at build 5599).

Section 1: Which Functions work in MediaCoder?

Table 1, below, indicates whether each of the tested functions ‘Works’ or ‘Fails’ for the builds listed or are succesful ('yes'). They relate to the 64-bit versions (x64) of MediaCoder and, thus, may not apply to 32-bit versions (x86)! In particular, please note that the tests were done under specific conditions using specified file types and Presets, described in Section 2. Thus, strictly speaking, they may apply only to these conditions, although I do expect them to apply more generally. They are meant to be indicative only and they may not apply to your special circumstance!

Compared to earlier tests, on the x86 versions of builds up to 5290, these tests have been simplified by removing some tests that had not failed for a long time. The deleted tests were Group, Audio Delay and Interlacing. Also, the test for Add Track B was modified and the tests ‘Effects A’ and ‘Effects B’ were combined. However, I have added a test to check that Xvidcore continues to work on large files (see above) and have modified the test for Hard Rendering.

Table 1: test results
Uploaded with ImageShack.us
* failed because all conversions failed (viewtopic.php?f=28&t=12404)
** see Section 3, below
# not tested but assumed to work

NOTE: before these comparative results may be correctly interpreted, they should be read in conjunction with Section 2, below.
If your build of MediaCoder fails in a test that is important for you, uninstall that build and install a build that works.
Earlier builds are available from http://www.videohelp.com/tools/MediaCod ... s#download.

Section 2: MediaCoder Functions & how I tested them

The functions in MediaCoder that I test (when I feel up to it), how I do it and what constitutes success are described below. Please note, though, that I had to restrict the extent of testing because there are too many perms and coms.
When a test requires the playing of an encoded output, I used VLC, a software media player.

Easy Cropper: (click on the Cropper button on the Picture tab).
The crop function is described in viewtopic.php?f=17&t=8188&start=0 and is considered working if the video can be moved 'under' the defined boundaries in Free Cropping mode and the boundaries may be dragged relative to the video for a different crop area. Usually when the Cropper fails it opens with no video showing (uniform grey).
The Cropper test is routinely done on a VOB file that has only one audio stream and no subtitle streams. An occasional test with an FLV file gives the same result.

Trim or Clip a video: (Clipping & Effects window via the 'Effects' button on the Picture tab or 'Select' from the Time tab).
'Clipping' is used to remove content from the ends of a file, and is described in viewtopic.php?f=17&t=8912&start=0. It should be reasonably easy to set the Mark In and Out points on this window. Whereas, trimming a video with such a tool would be inaccurate, depending on the I-frame spacing, it ought to be usable. The slider should stop on hitting Pause and it should be possible to scrub through the video to find the desired points. The jump buttons (double arrows) should also work (~10 sec jumps). The test result will be indicated as “fails” if both of the following methods fail: (1) move the slider repeatedly by dragging or the jump buttons to find the desired points and mark the times with the Mark In/Out buttons and (2) use the Pause button to allow scrubbing and marking. It was tested on a VOB file that had one audio stream and no subtitles.

Copy the Video stream: ('Copy Video' on the Video tab).
Presumably, this 'copy' function would fail if the video and/or audio streams were not compatible with the output container. For example, it would usually fail for an FLV file converted to MPEG2. Copy Video may be applied with the audio disabled, converted or copied. Similarly, there are three options for Copy Audio. If all combinations were to be tested, especially if I were equally thorough with the other functions, testing would take too long. It would be even more complicated if I considered combining two functions.

Video quality is easily compromised by repeated encoding, whereas, in my opinion, it is easy to re-encode audio without a noticeable loss. Moreover, the saving in time when video is copied is far greater than for copying audio. Hence, the ability to copy video will be considered as more important; and I will limit myself accordingly. Consequently, the 'Copy Video' function was tested by converting a VOB having just one video and one audio stream, and was considered to be 'working' if (1) both the video and the audio will copy at the same time (very quick since no transcoding takes place) and (2) Video will copy with the audio stream being converted. The test was done using the MPEG2 Preset since this container will accept the streams from a VOB (see NOTE at the end of Section 3).

Choose Audio Stream: (select ID on audio tab).
This was tested by applying the MPEG2 Preset (see NOTE at the end of Section 3) to a VOB file having three different audio streams. This feature is considered as 'working' if the selected audio stream appears in the output file when converting both the video and the audio streams of this VOB. I chose to limit myself to this scenario since there are too many possible combinations and I have had to restrict myself.

Hard Rendering of a Subtitle: on the Subtitle tab, choose Mode = '*** Render', since build 5335, or Render = 'Default' for early builds. See also the comments in Section I of the guide on rendering (viewtopic.php?f=17&t=10927).
For this test, I chose the above VOB with 3 audio streams and 8 subtitle streams and I converted it to MPEG2 and/or to MP4 with the ID = 0 (default, English) and with ID =7 (German). I also tried to hard render an MKV with one subtitle track when converted to MP4. For success in both tests, the correct subtitle had to be present on the Preview panel (if open) during transcoding and also in both output files.

However for rendering to be considered a successful function, the setting of 'Disable' should avoid rendering. Hence, if subtitles are present when required and absent for 'Disabled', the rendering function is considered as “working”. On the other hand, if rendering occurs when disabled, it is just as much a failure as not rendering when 'Render' is chosen. I shall not distinguish between the two. The result is expressed in the Table as “x/x” with the first ‘x’ referring to the VOB source and the second to the MKV, with x = w, if the test works, and x = f, if the test fails.

Add Track feature: (on the 'File' menu or Ctrl+T). This feature is described in viewtopic.php?f=17&t=8082&start=0.

Add Track A” test: a VIDEO_TS folder was used that had been ripped to a HDD from a DVD in PAL. It has one VTS group having 14 'Tracks' and MP2 audio. I chose a DVD with MP2 audio because a problem of this kind seems less likely for AC3 audio. There were three steps:
Step 1: use File > Add Track to select the VIDEO_TS folder, via Disc Root/URL, to display all its tracks in the Track panel and to select tracks 7 and 11 to display their content in the Chapter panel at right,
Step 2: 'Add' these tracks into MediaCoder and check their details in the Properties panel and
Step 3: convert Track 11 to MPEG2.
The Properties info (Step 2) was either satisfactory or mostly empty. When 'satisfactory', it had details under Container, Video and Audio, even though usually only broadly correct. When 'mostly empty' there were very few data and a conversion (step 3) usually failed.
This test was rated using three letters in the form: “A/B/C”, where A, B and C relate to Steps 1, 2 and 3, resp. Each is either 'w' if it works or 'f' for failure. A failure for step 1 is when it gives a complete blank in the Track panel or gibberish in the chapter panel; and for step 2, when the Properties data for either track is virtually empty. Obviously, if the Track panel is a blank (Step 1), a track could not be selected for adding to MediaCoder for conversion. In this case, the result would be “f/f/f”.

"Add Track B" test: I used a DVD in PAL with 12 VTS groups for 12 ‘Tracks’. I used Add Track to extract all 7 chapters of Track 4, as well as “Whole Track” and converted to MP4, as per the Encoding Guide. The test has two components: (1) to see if the chapters had missing frames (explained in the bug report viewtopic.php?f=28&t=12368) and (2) to check whether the output file had the correct Display Aspect Ratio when set as ‘Keep Display AR’ and 720x576 (see bug report: viewtopic.php?f=28&t=12366). A result of “w/f” signifies that the chapters were correct and the resultant DAR was not (or the conversion failed).

Video Effects: (applied via the Effects button on the Picture tab).
Here, there are two groups: (1) Brightness, Contrast, Saturation and Hue, and (2) the three colour gammas, nominally for Red (R), Green (G) and Blue (B). Notice that Hue is like a Colour Wheel, adjustable from -180 to +180 to cover the spectrum, with small values near 0 (default) offering small colour corrections.
Notice also that the success of each group of Effects is expressed separately in the Table of Section 1 as A/B, where A refers to the first group and B to the second (‘w’ for success and ‘f’ for a failure). For either group of effects to be considered successful, they must satisfy two criteria: (a) the effect of any adjustment must be evident in the video panel of the Clipping and Effects window, else this feature would have no practical use, and (b) that the change seen in this panel should be a true reflection of the change to be seen in the output file after conversion.

Encoding Test: I applied the various encoding suggestions tabulated in the Encoding Guide (viewtopic.php?f=17&t=9643). In each case, the source file was a 100-sec VOB file, which has one MPEG2 video stream at 6000 kbps, one AC3 audio at 224 kbps and no subtitle streams.

Xvid FourCC code: If the correct FourCC code is not written to the header of an Xvid/AVI file, many standalone players will not play. It should be ‘dx50’ or ‘XVID’, for example. For this test, the build is considered successful if “Codec ID = XVID” in MediaInfo for the Xvid/AVI settings given the Encoding Guide (viewtopic.php?f=17&t=9643). Otherwise it fails, usually with “FMP4” displayed. The failure may be overcome by including in your Preset: Settings > Overall > Video > “Override FourCC …” = XVID. Alternatively, change to a build that works.

Xvidcore: This test was to check whether the Xvid/AVI settings in the Encoding Guide, using the Xvidcore encoder, work on large source files. I used a 12min 51sec MP4 source file. By comparison, Xvidcore failed such a test in all 32-bit builds I had tried.

Section 3: More Observations

The Encoding Guide: the Encoding Guide (viewtopic.php?f=17&t=9643) was changed at various times:
(1) in build 5290, the audio for WMV was changed from ADPCM, used in earlier tests for x86 builds, to WMA-8.
(2) In builds 5342-48, Nero Encoder failed for sampling frequencies below 32 kHz and the bug was fixed in build 5350.
(3) Build 5370 failed to work with Presets (viewtopic.php?f=28&t=12546), which bug was improved in 5372. However, builds 5372 to 5380 fail to save the Output Folder to Presets (see: viewtopic.php?f=28&t=12579). This was rectified in the version of 5380 dated 3 Apr 2013.
(4) Build 5392 still has 3 Bugs that had been present a long time, as were noted in viewtopic.php?f=17&t=12627.
(5) Build 5399 - 5505 would not copy either the video or the audio (see: viewtopic.php?f=28&t=12643)
(6) Build 5505 failed to convert to MP4 via Add Track (viewtopic.php?f=28&t=12659) and failed encoding AVI to MP4 (viewtopic.php?f=28&t=12658)
(7) Build 5540 failed to convert whenever FFmpeg was used as Video Encoder (viewtopic.php?f=28&t=12747 )

NOTE: for the MPEG2 Preset, mentioned above, I used the settings given in viewtopic.php?f=17&t=8599&start=0, which includes the setting of the GOP length to 12.

Keep tuned for more instalments (especially updates to Section 1)!
Have you checked out the Tips & Guides for MediaCoder? Try: http://forum.mediacoderhq.com/viewtopic.php?f=17&t=8061
Also, get older builds at: http://www.videohelp.com/tools/MediaCoder/old-versions#download