PowerDirector 7 includes 10 effects that can leverage CUDA acceleration during rendering. If you’re new to editing, the process boils down to this: you take your video clip, put it on your editing timeline, add effects (blurring, sepia toning, or whatever), and when you go to output the project into your target format, the video clip and effects get rendered into a final file. This sounds simple, but applying effects to video is tremendously compute-intensive. In fact, rendering has traditionally been the bane of video editors because one render would consume a system’s resources for hours, bringing productivity to a standstill. The size of files may be different, but home editing falls prey to the same resource limitations.
Armed with a 30-second, 720x480 MPEG-2 test clip, we added the CUDA-assisted Pen Ink effect and rendered into an MPEG-2 output file. With CUDA enabled, our render time was less than half of putting the render load entirely on the CPU on both cards. But notice two things here. Unlike with SETI@home, we see little if any benefit from all those extra stream processors in the 9800 GTX. Additionally, employing CUDA only shaved 3% to 5% of the load from the CPU, which remained almost entirely consumed by the render job. In this test, CUDA will help accelerate your task, but you’re still going to have a buried processor unable to handle any other applications.
The second test set involved a larger 1080, H.264 clip from the HDNet show Get Out!, which we then exported into PowerDirector’s MPEG-2, AVCHD 720 x 480, and AVCHD 1920 x 1080 profiles. Again, we see very little difference in performance between the 9600 GT and 9800 GTX. Also note that there is no difference in exporting into MPEG-2 with or without GPU acceleration because NVIDIA’s library is only supporting H.264 encoding, not MPEG-2 encoding. The moral of that story is that if you’re in the habit of transcoding movies into MPEG-2 files for playback on a device that doesn’t support H.264, then CUDA will be about as useful to you as a third arm while jogging.
Still, you can see that CUDA gives us over a 100% improvement on the 480 test and a nearly 300% improvement when encoding to 1080. Extend this task to the duration of a movie and you can start to see how much time CUDA might save you and your system.