Bill Anderson wrote:
> I built my current system in August, 2007, and as time has gone by I've
> changed a few things so that I am currently running Windows 7 64-bit on:
>
> Motherboard: ASUS P5Q Pro Turbo
>
> Processor: Intel Core 2 Quad Q6600 Kentsfield 2.4GHz 2 x 4MB L2 Cache
> LGA 775
>
> Memory: Crucial 4GB Ballistix 240-pin DIMM, DDR2 PC2-8500
>
> Video Card: SAPPHIRE 100210L Radeon HD 2600XT 256MB
> 128-bit GDDR4 PCI Express x16
>
> Power Supply: PC Power and Cooling ULTRA-QUIET PSU:
> SILENCER(R) 610 EPS12V
>
> Video capture: ATI VisionTek TV Wonder PCI Express
>
> Storage: Four HDDs -- 1 Samsung 1000 GB 7200 RPM SATA II, 1 Seagate
> 1500 GB 7200 RPM SATA II, 2 Seagate 1500 GB 5900 RPM SATA II.
>
> For the past few days I've been authoring DVDs (which has involved
> transcoding video files) and it just seems to me that the process is
> taking far too long. When I run Adobe Encore my CPU usage gadget
> indicates that all four cores are constantly maxed-out for up to six
> hours straight, just to produce one DVD (ISO file). My current setup
> works -- in fact it works great -- but it just takes SO long to do the job.
>
> I know this sounds a lot like Allan's current query here on the
> newsgroup -- maybe there's answer to this question that will help both
> of us: Can I build a new computer (MBO, CPU, Memory, Video, HDDs) that
> will give me a significant, notable, impressive, breathtaking
> improvement in the processing time for video files? Thanks.
>
I think you'd be better off surfing a forum dedicated to the subject first.
Your rendering times may depend on the resolution of the content.
You can always find *someone* out there, with abnormally long render times
(and no explanation as to why). The cheapest fixes for performance, are
in the software. It's pretty difficult to fix every problem, with
faster hardware. (Poorly written software, trumps spiffy hardware every time.)
http://forums.adobe.com/message/2575261?tstart=0
Hardware solutions include
1) CPU
2) GPU (CUDA acceleration)
3) Third party accelerator cards, Matrox or Spurs Cell Engine or the like.
In some cases, there is a performance crossover. With weaker accelerator solutions,
they actually work slower than the CPU. Eventually, with enough horsepower, the
losses going with some kind of accelerator, eventually give you enough improvement,
to surpass a CPU-only solution.
You have to cost these options out, and see which ones make sense.
http://en.wikipedia.org/wiki/CUDA
http://en.wikipedia.org/wiki/SpursEngine
If you read the comments here, the SpursEngine may require a separate software
path to get the job done. (I.e. If it isn't integrated into Encore to begin
with, you may end up with a very screwy work flow.)
http://www.techpowerup.com/130454/Le...One_Board.html
And finally, a solution from Matrox (home of "expensive cards" :-) )
Now, the second page here, doesn't make it look that much faster.
http://www.matrox.com/video/en/products/pc/compresshd/
http://www.matrox.com/video/en/products/pc/max/media/
The Intel processors with QuickSync, can speed up processing to some extent,
as they can do things like have enough decoding power to decode five movies
at the same time. (5x realtime on the decode end of things).
The trick is, to find a worthy web page, that benchmarks stuff like this,
head to head (with same software settings in each case).
*******
If you didn't choose to do any more research, and just wanted to upgrade the
CPU, I won't argue with that logic. Software using CPU is the most flexible,
and the most likely to have fixes released for it.
You can compute the ratio of clock rates, on the core(s) of your new processor,
versus the old, and get some improvement. The purpose of me mentioning the other
solutions, is so you won't have "regrets" later, that you didn't look further
at other ways of doing it.
http://www.cpubenchmark.net/high_end_cpus.html
2600K passmark = 9977 price = $315 USD
Q6600 passmark = 2983
So a new processor, gives about 3x on that benchmark. If you absolutely must have
the best processor, Intel generally wants $1000 to $1500 or so, somewhere in that
ballpark. (More than that, you're looking at server motherboards and server processors.)
(Intel pricelist - see E7 series, for examples of "over the top" - 10 cores max)
http://files.shareholder.com/downloa..._1ku_Price.pdf
Then, it's a question of whether any other hardware solutions (CUDA or accelerator)
can surpass that, at a reasonable price. In the case of CUDA, there may not
be a simple proportionality between price and performance (maybe a midrange
card gives most of the performance of a high end card, depending on where
the bottlenecks are - a GTX260 might do just as well, as something newer).
Also, there can be minor quality differences between the various hardware
solutions. Obviously, a general purpose processor has the most opportunities
to have any issues fixed - with specialized solutions, you don't know how often
(or if ever), patches or improvements to the "microcode" will come out (whatever
passes for code inside the accelerator type solution).
Paul
Similar Threads
Linear Mode
