|
|
Posted by Neil Smith [MVP Digital Media] on 03/18/06 21:07
On 16 Mar 2006 10:39:47 -0800, "Scubajam" <jmcgauhey@usa.net> wrote:
<snip>
>the original HD). I went HD all the way, until final creation of mpg
>for DVD. I rendered multi-pass at 8000 VBR. Quality is excellent.
>Final file is 1.5 gig for 30 minutes. Burned on Taiyo Yuden media. I
>also rendered a HD file, but the only way to play back HD at this time
>is to Firewire back to camcorder and burn a HD MiniDV tape. Then must
>use camcorder as deck for playback into HD TV. It took 6 hours to
>render on a AMD 64 3000+ with 2G RAM, using multiple harddrives (one
<snip>
Since a computrs GPU scales far better for repetitive tasks like
encoding compared to a general purpose CPU, it can achieve really
large ramps in encoding speed.
Mostly that's down to the high level of parallelism the GPU offers in
its instruction pipelines, which you'd normally be using to render
Quake textures - the GPU isn't limited to just graphics though, and
could be re-purposed to assist or take over encoding tasks instead.
Interestingly, ATI recently released the Avivo encoder application. It
utilises the GPU instead of the CPU to do transcodes and encoding.
I'm due to test it here this weekend, but I expect other encoder
makers like MS and Apple would eventually release something which uses
the same technique. I've already seen a paper by MS research which
addresses using the GPU for tightly focused processing of batch tasks.
The http://www.gpgpu.org site deals extensively with these programs.
Some of these links might be of interest to you :
http://www.chip.de/artikel/c1_artikel_17670022.html?tid1=9227&tid2=0
http://www.extremetech.com/article2/0,1697,1880669,00.asp
http://forum.doom9.org/showthread.php?t=104157
http://research.microsoft.com/research/pubs/view.aspx?msr_tr_id=MSR-TR-2005-183
http://research.microsoft.com/research/pubs/view.aspx?type=technical%20report&id=1040
HTH
Cheers - Neil
Navigation:
[Reply to this message]
|