David Taylor
August 6th, 2008, 10:39 PM
It's about 50% faster than B58 on most source material. It also fixes many of the pesky gamma issues FCP/QT is known for.
http://www.cineform.com/products/Downloads/DownloadMacOSTrialStart.htm
One obvious question is for those who care - why is it so much faster? Well, our codec core shares the same code base with Windows which has been efficiently threaded for a very long time. But it turns out that FCP/QT do a lot of funky/odd things when requesting data from the codec. And because we have so many different modes: 8/10/12 bit, 4:2:2/4:4:4/Raw, and up to 4K spatial resolution, the combinations and permutations of different data requests to the codec are unpredictable. It turns out that each time FCP/QT makes an "odd" call it was dropping into single-threaded code. that killed the performance of our otherwise well-threaded codec. The good news is that we've figured out most of these odd data requests and have threaded the associated code.
There is one more routine that will be threaded in the next release (probably within a few days or so). But our Mac performance is now getting to the point where it is quite respectable.
If you haven't tried it for a while, give it a try. It also comes with our Active Metadata application for Mac which is very powerful. NOTE: If you create files on Mac to use with Active Metadata, they must be created with Build 56 or later.
Have fun....
http://www.cineform.com/products/Downloads/DownloadMacOSTrialStart.htm
One obvious question is for those who care - why is it so much faster? Well, our codec core shares the same code base with Windows which has been efficiently threaded for a very long time. But it turns out that FCP/QT do a lot of funky/odd things when requesting data from the codec. And because we have so many different modes: 8/10/12 bit, 4:2:2/4:4:4/Raw, and up to 4K spatial resolution, the combinations and permutations of different data requests to the codec are unpredictable. It turns out that each time FCP/QT makes an "odd" call it was dropping into single-threaded code. that killed the performance of our otherwise well-threaded codec. The good news is that we've figured out most of these odd data requests and have threaded the associated code.
There is one more routine that will be threaded in the next release (probably within a few days or so). But our Mac performance is now getting to the point where it is quite respectable.
If you haven't tried it for a while, give it a try. It also comes with our Active Metadata application for Mac which is very powerful. NOTE: If you create files on Mac to use with Active Metadata, they must be created with Build 56 or later.
Have fun....