[FFmpeg-devel] transcoding on nvidia tesla

[Måns Rullgård]

Mike Melanson <mike at multimedia.cx> writes:

> M?ns Rullg?rd wrote:
>> "Dan Khasis" <dan at novapulsar.com> writes:
>> 
>>> I can make a significant contribution to get this going, and
>>> purchase a card or two for development testing.
>>>
>>> When it comes to converting thousands of videos per second/minute
>>> on one or two teslas, versus 10,000 servers across 10 datacenters,
>>> using kilowatts of power, the costs associated with several of
>>> these servers become irrelevant.
>>>
>>> I guess the real question is if it's possible, how long it would take, and
>>> how much I would have to contribute :)
>>>
>>> Based on my understanding of how well the developers of ffmpeg, have made
>>> ffmpeg the de facto video conversion application, I think it's safe to say
>>> they can do this too.
>>>
>>> With a tesla or two, you could be transcoding thousands of HD streams
>>> simultaneously.
>> 
>> I doubt the performance boost would be as big as you suggest.
>
> I think the belief -- partially driven by the marketing material -- is
> that, since one of these cards has 128 thread processors, it should be
> able to run 128 concurrent instances of FFmpeg and encode 128 HD videos
> simultaneously. There must be some parts of, e.g., the H.264 encoding
> process (which, as we know, does not live in FFmpeg) that could take
> advantage of parallelization. Of course, there's still the overhead of
> little matters like bitstream syntax encoding and file muxing.

Is there any reason to believe that each of these threads has the
power of a full CPU at its disposal?

-- 
M?ns Rullg?rd
mans at mansr.com