|
Imagine 128 S2 vs. Twin Turbo M8 - Movie Playback Tests
By Mike, Published: 8/17/1997 |
|
| Movie Playback Performance Testing | |
|
Full screen playback of scaled movie files is a tough test of a cards scaling
and pixel interpolation engines. These tests were run at at 1024x768, thousands
colors mode for the 16-bit tests, and 1152x870, millions color modes for the hires tests.
The results at 16-bit color (thousands) settings is almost a draw. My initial
feeling was that 1024x768 was a good starting point, but I expected to see more of a difference between
the different cards. Using a CPU card with accelerated bus speeds really benefits the 8500 video -
another example where an CPU accelerator card is an excellent value for more than just cpu
speeds. With the 50mhz bus speed the 8500 does quite well in the thousand colors modes. Be aware only the Imagine 128 has a 256-bit Video engine, and a recent MacWeek article cited: "With a Cinepak-compressed QuickTime movie, Imagine scaled smoothly to 1,600 by 1,200 pixels; TwinTurbo dropped frames even at 1,152 by 870 pixels.". The Imagine again appears to have a higher performance design than the competition, but for resolutions up to 1024x768 all cards in the test (including the 8500 video) did reasonably well. We used MacBench's "custom" video test to define the parameters and source file for the full screen tests, and selected their standard Cinepak2 movie file. A cinepak2 320x240 movie was used and set be scaled to full screen. I used the "copy to ram" option to eliminate any disk influence on the scores. The results are summarized below in precentage of frames played and max frames per second (fps).
Fullscreen Video Test: 1024x768, thousands colors, 75hz vertical
refresh As shown in the graph above, even the built-in video of the 8500 did a fairly good job at this bit depth and size, and the image quality was surprisingly good, but the Imagine 128 and Twin Turbo images were better quality than the built-in video when movies were scaled to full screen size at high resolutions. This next graph shows the results of the full series of MacBench video playback tests which display the percentage of frames played and the maximum framerate for a variety of video formats and sizes. Note that all cards had problems with motion jpeg and uncompressed video files - I suspected the 4X cdrom in some cases but I'm sure MacBench would copy the files to ram to eliminate the CD speed affecting the score.
As a test to see if the Imagine 128 would distinguish itself from the pack at higher color depths and resolutions in video playback as it had in the graphics tests - I reran the custom video test in true-color mode (1152x870, millions colors). As I had come to expect, the Imagine's performance showed the least performance drop. The copy to ram failed at this resolution, and had to be run from the CDrom. The graph below shows the results:
Full Screen Video: 1152x870, millions colors (ran from CD): As I had not run the 8500 video in this test - I decided to rerun the test with the 8500. This time I created a 11 meg Ram Disk to hold the video. The Twin Turbo scored much better running the movie from a ram disk. I find this odd, and have no explanation for this. A graph of the same test as above run from a ram disk is below:
Full Screen Video: 1152x870, millions colors (ran from Ramdisk): As you can see in the video tests are almost a dead heat at 16-bit color modes, but in true-color the Imagine 128 again shows a slight advantage. I'm not sure that's a big decision point for most buyers however, as 1024x768, 16-bit color movies are good enough for most users. The higher res/color tests do show that the Imagine engine is slightly better in performance however. Video Test Winner: Imagine 128 (practically a draw however) |
|
Brand/product names mentioned here are properties of their respective companies. Legal: Users of this web site must read & are bound by the terms & conditions of use. For more recent video card reviews - see www.xlr8yourmac.com/video.html. |
|