From: ldo@waikato.ac.nz (Lawrence D'Oliveiro, Waikato University)
Subject: QuickTime performance (was Re: Rumours about 3DO ???)
Organization: University of Waikato, Hamilton, New Zealand
Lines: 67

OK, with all the discussion about observed playback speeds with QuickTime,
the effects of scaling and so on, I thought I'd do some more tests.

First of all, I felt that my original speed test was perhaps less than
realistic. The movie I had been using only had 18 frames in it (it was a
version of the very first movie I created with the Compact Video compressor).
I decided something a little longer would give closer to real-world results
(for better or for worse).

I pulled out a copy of "2001: A Space Odyssey" that I had recorded off TV
a while back. About fifteen minutes into the movie, there's a sequence where
the Earth shuttle is approaching the space station. Specifically, I digitized
a portion of about 30 seconds' duration, zooming in on the rotating space
station. I figured this would give a reasonable amount of movement between
frames. To increase the differences between frames, I digitized it at only
5 frames per second, to give a total of 171 frames.

I captured the raw footage at a resolution of 384*288 pixels with the Spigot
card in my Centris 650 (quarter-size resolution from a PAL source). I then
imported it into Premiere and put it through the Compact Video compressor,
keeping the 5 fps frame rate. I created two versions of the movie: one scaled
to 320*240 resolution, the other at 160*120 resolution. I used the default
"2.00" quality setting in Premiere 2.0.1, and specified a key frame every ten
frames.

I then ran the 320*240 movie through the same "Raw Speed Test" program I used
for the results I'd been reporting earlier.

Result: a playback rate of over 45 frames per second.

That's right, I was getting a much higher result than with that first short
test movie. Just for fun, I copied the 320*240 movie to my external hard
disk (a Quantum LP105S), and ran it from there. This time the playback rate
was only about 35 frames per second. Obviously the 230MB internal hard disk
(also a Quantum) is a significant contributor to the speed of playback.

I modified my speed test program to allow the specification of optional
scaling factors, and tried playing back the 160*120 movie scaled to 320*240
size. This time the playback speed was over 60 fps. Clearly, the poster who
observed poor performance on scaled playback was seeing QuickTime 1.0 in
action, not 1.5. I'd try my tests with QuickTime 1.0, but I don't think it's
entirely compatible with my Centris and System 7.1...

Unscaled, the playback rate for the 160*120 movie was over 100 fps.

The other thing I tried was saving versions of the 320*240 movie with
"preferred" playback rates greater than 1.0, and seeing how well they played
from within MoviePlayer (ie with QuickTime's normal synchronized playback).
A preferred rate of 9.0 (=> 45 fps) didn't work too well: the playback was
very jerky. Compare this with the raw speed test, which achieved 45 fps with
ease. I can't believe that QuickTime's synchronization code would add this
much overhead: I think the slowdown was coming from the Mac system's task
switching.

A preferred rate of 7.0 (=> 35 fps) seemed to work fine: I couldn't see
any evidence of stutter. At 8.0 (=> 40 fps) I *think* I could see slight
stutter, but with four key frames every second, it was hard to tell.

I guess I could try recreating the movies with a longer interval between the
key frames, to make the stutter more noticeable. Of course, this will also
improve the compression slightly, which should speed up the playback performance
even more...

Lawrence D'Oliveiro                       fone: +64-7-856-2889
Computer Services Dept                     fax: +64-7-838-4066
University of Waikato            electric mail: ldo@waikato.ac.nz
Hamilton, New Zealand    37^ 47' 26" S, 175^ 19' 7" E, GMT+12:00