Originally Posted by digitalclips
Indeed as we get these staggeringly high download speeds it all to often reveals the limitations of the other end. On the other hand I started watching the NBC Live Olympics on 75 Mbs download at home and went on vacation to a place with 5 Mbs and there was zero difference in the HD quality which truly shocked me.
That's because "HD" on Cable is either a MPEG-2 TS with H262 video or a MPEG-2 TS with a h263 or h264 video stream. Seriously, I can check what my cable box spews out right now...
Video: MPEG2 Video 1920x1080 29.97fps 24000kbps [Video - MPEG2, Main Profile, High Level, 1920x1080, 29.970 fps, 24.00 mbit/s (0800,e0,00)]
Audio: Dolby AC3 48000Hz 6ch 384kbps [Audio - English, Dolby Digital, 48.0 kHz, 6 chn, 384.0 kbit/s (0801,bd,00)]
I've had the DSL version before as well, but they do it with Mpeg2 TS with h264. Running two SD streams takes 5Mbits. I can't claim anything about the HD as I don't live there anymore.
The trick in video compression is that it scales linearly.
MPEG1 ran on a 386 with a 1X CD-ROM
MPEG2 ran on a Pentium with a DVD-ROM
Then we had this lull where Divx/xvid jacked h263 and it became the pirate codec of choice until they ran into file size limits. Mobile phones still were in dumbphone phase until at least 2005 when camera phones started being all the rage and MMS messages (remember those?) started happening. However it was sites like Youtube that actually started utilizing the codec with the flash plugin. Prior to youtube, all video sites were terrible.
However h264 and the x264 codec is what runs sites like Youtube, It's also the native video format used by all consumer cell phones, camcorders, pictures cameras, webcams, animation software, etc. So this is what we're stuck with until h265. Perhaps maybe something reasonable can be decided with h265 that all hardware and software can encode, and decode freely, and limit licencing to hardware (not software.) Software, it's inevitable that an x264-like project will come along, perhaps written natively against openCL so that an encoder and decoder can work on the underlying GPU hardware. The reason there are "hardware accelerated" GPU's for mpeg codec's is largely because GPU's used to be fixed function. This hasn't been that way for quite a while, so it's certainly possible to accelerate a h265 video on current video hardware. Encoding however will still require more powerful hardware.
As it is, current h26x video encoding doesn't make efficient use of multiple cores in a CPU, or multiple GPU's because the compression process is still linear. What needs to happen is developing a codec that scales with available bandwidth and processors. If a system is incapable of decoding a 2560x1400 video, it should downgrade it to 1080p, 720p, 480p, etc so that it still plays smoothly. Currently video players just drop frames which leads to ghosting or motion-blur effects. Devices with smaller screens wouldn't need to scale up, just scale down.