Sign in with
Sign up | Sign in
Your question

dual core and avi transcoding

Last response: in CPUs
Share
May 25, 2005 9:25:24 AM

right, been doing a little reading up on dual cores and what they are promising and i've gone and gotten myself all confused (not that that takes much mind...)

anywhoo, the question is (and eden i'm hoping you in particular can shed some light on this one) will dual core actually benefit me to any great extent in avi-mpeg2 transcoding operations?

i had initially expected that it would lead to much faster transcoding operations and thus would reduce my working time on any one project. however, the jist i am now getting is that it won't actually make the rendering any quicker but will allow me to continue to use the computer for other operations whilst the rendering is in progress without slowing down the rendering process. have i got this arseways or am i on the right track?

________________________________________________________
know your limits, and never stop trying to exceed them!
May 25, 2005 9:50:44 PM

>Currently those encoders are still very much designed with
>dual chip in mind for multithreading as opposed to dual core.

Oh yes, and we all know how different it is to code for an SMP system as opposed to a dual core system. Its quite a miracle really all that old SMP software even works on dual core, is it not ?

note to the original poster: rendering and audio/video encoding are probably the most ideal applications to benefit from multicore. Most apps/codecs are ready, and the workload is simply ideally suited to a parallel approach, meaning, unlike with most other apps, you might see close to a 2x speedup over a single core chip at the same speed.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Related resources
May 26, 2005 7:01:25 AM

>Go read Anandtechs article about cache communication latency
>and how the two cores interact to the memory controller. It's
> those parts that can be futher optimised for dual core.

I had read the article, and pretty much knew what was written there anyway, but no where does it say software can or will be optimized to benefit K8s fast cache to cache transfers; any threaded software that requires lots of intercpu communication already does (but not many do, as there is barely any app showing a measurable performance increase between 2xopteron and 1xdual core opteron in spite of the massive latency advantage).

If anything, optimizing threaded software includes *avoiding* c2c as much as possible.

Besides, this is not a feature of dual core chips, its a feature of K8s. Pentium Ds c2c bandwith and latency is identical to a 2 way xeon since it lacks a crossbar and goes over the FSB.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
May 26, 2005 7:06:54 AM

>Oh good, that makes my AthlonMP server well spent!

Might be, but for sure it wouldn't be much worse than a theoretical dual core athlon MP at rendering. And even more sure, software optimizations wouldn't change that :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
May 26, 2005 9:36:58 AM

ok, so if i were considering building another machine *specifically* for converting 30GB+ AVI's to MPEG-2 on DVD, is dual core or dual chip the way to go?

should i consider something alone the lines of the thunderk8we with 2x dual core opterons?

also, in the past i have always thought that the GC has no impact on processes such as video rendering - am i right here, or does a powerful GC help (6800, x800, etc...) ?

________________________________________________________
know your limits, and never stop trying to exceed them!
May 26, 2005 9:48:51 AM

>ok, so if i were considering building another machine
>*specifically* for converting 30GB+ AVI's to MPEG-2 on DVD,
>is dual core or dual chip the way to go?

Either of them. There is no significant (if any) performance difference between multicore and multicpu with equal number of cores and equal clock for these workloads, so let price/performance guide you.

>should i consider something alone the lines of the
>thunderk8we with 2x dual core opterons?

If 2x dual core fits your budget, by all means, get one. Not much point in paying extra for a board with 2x PCIe though, but I guess that price difference will not really be felt in the overall cost. Still, if its only for videotranscoding, might as well look for a cheaper dual board.

Also, 2x dual core opteron is going to cost an arm and a leg. Again, if that is not an issue, and performance is more important, get one. If not, if its for personal and more occasional use, sticking to just 2 cores would make more sense I think. In that case, wait until june, and consider an Athlon X2 or even Pentium D. These will be vastly cheaper, and still offer considerable video encoding performance.

>also, in the past i have always thought that the GC has no
>impact on processes such as video rendering - am i right
>here, or does a powerful GC help (6800, x800, etc...) ?

Nope, video encoding will work just as fast (/slow) with a 3D Rage videocard.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
May 26, 2005 12:03:29 PM

nice one.

the unit is for a studio which is running 4-5 evenings of the week (wedding videos are my own little sideline business) so if i can cut down my rendering times i can increase throughput by not having the machine tied up as long. i won't be building the new machine before sept/oct time as the P4 3.0 sys i have is doing the job steadily and reliable, it'd just be nice to be able to get the trancoding done in less than real time (right now it takes about 180mins for a 140min project), so i probably will wait and use Athlon 64 X2's if i go for dual chip

________________________________________________________
know your limits, and never stop trying to exceed them!
!