have found in Loiloskope for fun times rendered a short movie:
With the CPU to MP4 5:46 min With CUDA yet exactly 1:57 min
What abstained from the other nonlinear editing systems on? The technique is still of NVidia kept ready for a long time.
Antwort von Georg:
CyberLink PowerDirector supports CUDA
Antwort von hannes:
True, but the "famous" but do not. Even what is one of Canopus or Corel? I do not.
Antwort von Jörg:
Quote:
Even what is one of Canopus or Corel?
blu is offered as long the Canopus FireCoder and sold for real money, should the interest there for an internal solution very, very be low ....
Antwort von domain:
What I still is not clear, supports CUDA really synonymous to the timeline operations themselves, or the speed increase relates only to the final rendering process?
Antwort von pailes:
"hannes" wrote:
What abstained from the other nonlinear editing systems on? The technique is still of NVidia kept ready for a long time.
A switch to an API such as Cuda is far from trivial. I think many Manufacturers are experimenting with these libraries, but how efficient the result is synonymous depends on what condition is the existing rendering code of a video editing solution. A change has not happened within a couple of days, that can already take a few more months until everything has been tightened. And NVidia Cuda is a pure technology, owners of ATI cards then look into the tube.
With a little luck we will be with OpenCL (; can get http://de.wikipedia.org/wiki/OpenCL) A hardware-independent solution to put the software Manufacturer.
"domain" wrote:
What I still is not clear, supports CUDA really synonymous to the timeline operations themselves, or the speed increase relates only to the final rendering process?
What do you mean with "Time Line Operation"?
Antwort von domain:
Well stop everything when cutting so obtained include real-time preview. Is this because synonymous with CUDA accelerated?
Antwort von Mylenium:
"hannes" wrote:
What abstained from the other nonlinear editing systems on? The technique is still of NVidia kept ready for a long time.
Öhm the time of the CUDA-Beta and the release of the v1 there was just 3 of the GeForce 7xxx series graphics cards or something, which could (in theory), but in practice due to buggy drivers but could not use all the features operational. Only with the 8xxx series went off the right. GPU code is also quite different knit than CPU code that makes it so not even on the go. And even then is still a question whether it makes sense. Very complex constructs that only s.einer very high number of shader units to function, so that only on expensive cards, which use to anyone. Since the effort in development in Comparison to the potential sales of such cards is too great. And naja - GPU code is still largely single-precision. Even for MPEG encoding you have to go there if in doubt, even compromise. In addition, and that its indeed the problem, it must always be present yet another variant that functions as a "software only" on the CPU, if the target system does not support CUDA. But questions as necessary security mechanisms in order to shoot the calculator does not always require skill and care and even some ...
Mylenium
Antwort von Mylenium:
"domain" wrote:
Well stop everything when cutting so obtained include real-time preview. Is this because synonymous with CUDA accelerated?
So when Loilo there is nothing like this and in other programs we should see this. But the question is rather academic in nature decode - For pure play's other factors are more important than the processor power on. The funzt yes even at 400 euros Billigrechnern now.
Mylenium
Antwort von jazzy_d:
The h.264 cards of Matrox and Canopus are indeed synonymous "only" for the final export.
Antwort von domain:
"Mylenium" wrote:
But the question is rather academic in nature - for a pure play's other factors are more important than the processor power
In a nonlinear editing this issue is not purely academic but practical. Whether the final rendering is in fact accelerating me is relatively no preference. To accelerate the timeline operations, especially under AVCHD, it is mostly.
Antwort von WoWu:
The answer is: yes The problem is in CPU architectures that a GoP has been loaded into main memory, processed, and deleted. The MPEG2 Gops but had only an I-frame. The I-frame only had to be kept so long in the memory until the GoP was processed. H.264 but has up to 16 I-Frames and GoP length of such 90th This means that it not only more storage space for the 16 I-frames are, but they still synonymous (and instead of 1 / 2 sec 4 sec now) must be kept in memory. Also, the volume of reference requests is increased accordingly. Another factor is that synonymous B-frames (; slices) are used as reference for motion compensation, which was previously not possible. Therefore, the fundamental difference between P-and B-slices only in style, lies in the encoded macro blocks. In B slices, the Prädiktionssignal from the weighted mean of two different Prädiktionswerte can be formed. B-slices to use two different reference picture buffer, a fact which makes considerable demands s.The computer architecture. All this, and a lot more running from the DSP and additional synonymous changes, of course, the behavior of a real-oriented composition. So far have come with the camcorders used codecs are only 4 I-frames used (; synonymous with regard to the still weak computer architecture), and thereby naturally synonymous not exhausted the possibilities of the codec. That should change, however, synonymous with the changing architectures.
