Video Acceleration and power use when playing video.
Ian Stirling
OpenMoko at mauve.plus.com
Sun Jul 8 15:34:43 CEST 2007
Nils Faerber wrote:
> Ian Stirling schrieb:
>
>>As I understand it, the chip under consideration will not decode audio.
>
>
> Well, this does not matter that much.
>
>
>>This essentially means that the CPU can not usually be slowed down, as
>>it is required to decode MP3 audio or many other formats.
>
>
> Why not?
> Decoding audio needs lot less CPU resources than video so I would
> estimate that the CPU can be slowed significantly. But even then there
> are other problems that migh prevent CPU slow down like hvaing to feed
> the decoder with data, which usually is done through the CPU which needs
> to read the data from some storage, probably demux the stream and pipe
> it into the decoder.
>
>
>>This basically means that there is very limited power saving if any
>>possible, compared to simply doing it all in the CPU.
>
>
> The CPU will use a lot less cycles for decoding the video. If power
> saving is your concern then it might as well help.
It doesn't really.
For this CPU, you really need to get it to slow mode - 12Mhz - to get
significant power savings.
Simply idling doesn't actually save much.
>>Of course, it will make more video able to be decoded, without transcoding.
>
>
> Right, and with much higher quality.
>
>
>>On some personal tests, I found on a 2.8" screen, mpeg1 video at 120*160
>> and 300kbits/s worked just fine, which means that the only benefit this
>>may have in terms of playback is if you want to have more than ~5
>>hours/gigabyte, using more advanced video encodings that the CPU can't
>>do, or larger video than is actually visually needed, for playback on
>>other devices.
>
>
> Uhh... 120*160 is pretty bad resolution :(
> With an active decoder this could quite easily be QVGA or even full VGA
> at full framerate. And this *is* a benefit IMHO...
On a 2.8" display?
How close do you plan to hold the screen?
More information about the neo1973-hardware
mailing list