CoreAVC is a small company that claims that they have the fastest h.264 decoder in the world. You have to see it to believe it. I had heard of the claims before, but I didn’t put much thought in it. But it’s indeed true.
Where Quicktime fails miserably, ffdshow can’t keep up, and VLC all but freezes, CoreAVC shines through. I tried their ‘CoreAVC Pro’ trial version today, and is able to deliver 30 frames per second at high quality of a high profile h.264 1920×1080/30p video, without sweating. It is SCARY to see this decoder only using 60% of my Hyperthreaded Pentium-4 3Ghz CPU, a four years old CPU. The decoder is worth every penny of the $15 it costs. And its newer version that’s not out yet, Enterprise, is supposed to be even faster as it will have full GPU support.
In contrast, Quicktime offered me something like 3-5 fps, ffdshow’s x264 decoder did about 10 fps (or about 15 if you tell it to drop quality), and VLC would have artifacts if I told it to skip frames and drop quality, while it was unwatchable and freeze-prone when I told it to use the default behavior (not to skip frames and use max quality that is).
You see, all this time, I had to use the “drop quality” feature on my MediaPlayerClassic (with the latest ffdshow-tryout versions) in order to be able to test my own 720p HD videos or enjoy other’s videos that I download from Vimeo. And even then smoothness was not there. I wouldn’t even think loading full 1080p h.264 videos in that PC, as the performance was terrible. But with CoreAVC, it’s a piece of cake. Now my h.264 1080p videos are as fast as the XViD, WMV and mpeg2 1080i/p performance on that PC — if not better.
Note: the test video was encoded at 1920×1080 with the x264 encoder and ffmpeg, 29.97 fps, AAC 128 kbps 48000 Hz stereo, 12 mbps average, 16 mbps peak, in the mp4 container. x264 specific optimizations: -me_method umh -subq 5 -coder 1 -trellis 1 -g 300 -qmin 10 -qmax 51 -qdiff 4 -level 41 -rc_eq “blurCplx^(1-qComp)”