3 weeks ago
I'm using the 1.4 beta of Fusion Studio desktop for Windows (188.8.131.522 [a5ab95fc]). I'm running on a Windows 10 Home system with an Intel Core i7-6700K CPU @ 4.01 GHz, 32 GB RAM, an NVIDIA GeForce GTX 1070, a SSHD system drive plus an eSATA 2 GB hard drive. I've got at least 50 GB of disk space on the SSHD system drive, plus more than 1 TB on the hard drive.
When I try to render a 7 minute video from my Fusion on my desktop it gets to about 30% done, memory is fully consumed, and then it crashes (exits without any dialog or warning... I don't get any errors from Windows either). I've watched system resource usage using the Task Manager during a render. I can only render if I break a long video up into sub-2:30 minute clips. Then, memory is consumed to about the 90% mark during the rendering/stitching of each clip, then it is released before the next clip gets stitched. I can also see that the rendering output is consuming 100% of the capability of the destination drive (my source is on SSHD, my output is on the eSATA). Rendering seems to take about a second for every 2.5 seconds of video (5.2K/30 FPS down to 4K).
I don't understand the software design decision behind holding all of the video information in memory while stitching instead of reading in, say, 30 seconds at a time, stitching it, storing it in a temp file on the drive, repeating until all of the video has been stitched, and then assembling the final video from the temp files. That would take a lot less memory and because of the lack of thrashing due to VM page swaps and the lack of the constraint of the output drive the performance should be a lot better.
Shouldn't I be able to readily render/stitch any video I capture on the Fusion with my system?
I love the promise of the Fusion, but the software needs some rearchitecting if it is going to live up to the hardware. Unless I'm doing something wrong, that is.