The InterMark video test, devised by NSTL, tests the video capabilities of an adapter by forcing it to play video at a consistent speed and monitoring the number of dropped frames. We tested a representative cross-section of chip sets from our test-bed of PCI graphics adapters by testing one board for each chip set. We played 30-frame-per-second video clips under the Indeo and Cinepak compression formats at normal, X1.5, and X2 resolutions.
We predicted that adapters equipped with video-accelerator chips would have a clear advantage in this test; this proved to be the case. The VideoLogic PCI Movie accelerator and its sidekick, the PowerPlay 964 video-accelerator chip, had significantly fewer dropped frames than the competition as resolution increased. At normal (i.e., X1) resolution, the PCI Movie performed more slowly than the Focus Tr
ueSpeed W32 PCI, the Diamond Stealth 64 VRAM, and the Orchid Kelvin 64 PCI cards. But, as the charts show, there was little to compare to the PCI Movie at X1.5 and X2 resolutions. In fact, it dropped fewer frames at X2 resolution than it did at X1.
The largest differences between the PCI Movie and the other adapters manifested themselves at X1.5 resolution, where a board with hardware pixel interpolation should have the biggest advantage. A hardware pixel interpolator allows videos to be scaled during playback without causing more work for the host processor. Pixel interpolation is the rough averaging of pixels between the original pixels when the resolution is stretched beyond the normal X1. The space between the normally mapped pixels is increased, and the adapter must make a decision as to what color pixel to place there. (See the hardware review ``Video for Free''.)
A board with pixel interpolation will average the value of the surrounding area and put an averaged value in between. This proc
ess results in shadings between sections of an image. A board without pixel interpolation will put in one of the pixels from the surrounding area, creating a blocky effect. For complex video streams, the additional work of scaling the image via software results in dropped frames and jerky playback. A hardware pixel interpolator may also result in higher-quality playback, as the video driver typically uses faster techniques (e.g., pixel replication) at the expense of image quality.
The PCI Movie has hardware pixel interpolation; it dropped only 68 frames at X1.5 resolution, while its closest competitor, the Orchid Kelvin 64 PCI card, dropped 118 frames. This is not to say that the PCI Movie offers cinema-quality video, but it provides better-quality PC video than the other adapters that we tested.
Motion Video Test in Two Formats
illustration_link (13 Kbytes)
For video tests, we selecte
d one adapter to represent each chip set in this Lab Report. We forced each clip to run at 30 frames per second at normal, X1.5, and X2 resolutions using Indeo and Cinepak compression formats. Unlike most of our tests, lower numbers indicate better performance (i.e., fewer dropped frames).
Smooth Movies
