Quicktime Playback Tests - PC vs Mac, 1.3.1f24 vs 1.3.0f06
I have been building a show in which I have been running the pre-release on W7.
I have been experiencing some performance issues so I decided to test a few scenarios..including a Mac. I thought I would share my test results with you.
For the show, I will be projecting a series of single clips that will require smooth playback and the ability to vari-speed and scrub to achieve sync with the performance. The scrubbing disqualifies using the Windows Direct Show playback method as it does not allow for speed changes or scrubbing.
I am running my project at 60fps.
I tested across three different machines and with both 1.3.1f24 ( I'll call it 'PRE') and 1.3.0f06 (I'll call it 'FULL').
My results were a bit surprising.
Clip: 640X360 PJPEG Quicktime @75% Quality
A) ASUS G75VX 2.4 Quad, 32GB Ram, 670MX w 3gb, SSD, W7 PRO
B) ASUS G750JH 2.4 Quad, 32GB Ram, 780M w 4gb, SSD, W7 PRO
C) iMac (late 2013) 27" 3.2, 24 GB Ram, 770M w 1gb, SSD, 10.9
FPS CYCLES VPO
**A PRE ** 59.5 129 43
**A FULL ** 60 449 10
**B PRE ** 59.5 132 10
**B FULL ** 60 451 8.2
**C PRE CPU ** 59.9 170-514 3-4.5
**C PRE CI ** 33-59.9 170-430 3-7.6
C FULL CPU 47-59.9 200-516 .5-4.5
**C FULL CI ** 35-59.9 199-480 1-6.6
I have been considering switching to Mac OS for QC and CI capabilities. In addition, I have been running the pre-release as I assumed it would be quicker.
Based on my tests, it would appear that the ASUS G750JH with W7 Pro running the Full Version does the best job of video playback.
I am wondering if anyone else has some pre-release vs full-version tests they can share, or any technical insights that might help me understand the substantial performance differences between the pre-release and the full version.
Also, I am curious about why the Mac playback performance varies over time while the Windows performance remains constant.
Dear Justin,Thanks so much for these detailed results. Can you post the patch you used for testing? Was it just movie players? Or?One thing: the Cycles calculation in 1.3.0f24 was actually not being done correctly. So the with regard to the cycles, the values cannot be compared between 1.3.0f24 and 1.3.1f06\. Also, I have an internal version (beyond the pre-release) that will give even better performance under Windows. If you'd like to test, I can share that with you if you PM me.Also, I am now recommending to all Windows users that you enable "Windows Native Playback" in the General tab of the Preferences, and start using WMV/AVI movies.However there are some things that won't work as well as QuickTime. Specifically, changing the 'play start' and 'play length' parameters interactively and scrubbing the video using the 'position' are not very good with WMV/AVI. Also, the Windows playback system does (DirectShow) not offer reverse playback (i.e., speeds less than 0) at all. But in terms of pure performance, WMV/AVI are much better.The pre-release shouldn't suffer in comparison to the current release; but I want to make sure that this is the case.Best Wishes,Mark
Very interesting @jtstep
I am working on windows, and have found drastic variance in playback speeds.
Based on your clip size and codec, the only system variables I think may have such an effect are:
-video card model.. they really do change drastically.. (I am running a similar win7 machine with a nvidia 460, if you supply the files I am happy to add to the system)
-additional system software installed. (I have found some system services seem to slow me down.. eg: apache/mysql etc..)
The newest version Mark has should clear up all playback issues for you.. I suggest you give it a try.
hi mark, I was testing with just a movie player and a projector. thanks for your response. this certainly clarifies the cycles count discrepancy between the pre release and the full version. I will pm you regarding the current version. I do have a question for you about direct show. is it possible to interactively varispeed in the current version with direct show? I am using small live speed adjustments with my clips to maintain sync with the performance. QuickTime does this flawlessly, but I found that it is not possible to do this in the pre release when using wmv files. please let me know. thanks. justin.
Dear Justin,Yes, it is probably true that DirectShow is not so nice when it comes to speed changes. That's one of the things QuickTime does well.Best Wishes,Mark
In my testing, DirectShow does not allow for any speed changes. Perhaps I am not using it correctly?
@Dusx, I wonder if system services would explain the big discrepancy between the G75VX and the G750JH. The video card on the JH is very very quick, but I did not think that it would make much difference when playing back quicktime movies.
I did do a few initial tests with WMV files using Windows Video 9 Advanced and the LAV decoders to see if I could get the GPU involved and it seemed to work. That is as far as I got as I was not able to get the movie player to varispeed the clips when using direct show.
Here is a link to the test clip. (It'll be live for only a couple days) It is a small 640X360 quicktime with PJPEG compression.
What AVI/WMV codec do you use? What decoder do you use?
Dear Jsteph,Did forgot the test clip link? I don't see it.Best Wishes,Mark
Ahhh, yes. Here it is.
I did test the varispeed again using WMV files and I cannot get it to play at anything but 100%.