View Full Version : Max compositing performance, movie files or numbered stills?

03 March 2003, 11:28 PM

For compositing work, is it preferable to have the clips stored as continuous movie files (like avi) or as numbered stills? The reason for asking is that a continous large movie file would likely benefit more from the increased STR of a stripe-set, than lots of small files would. Is this correct? The clips are non-compressed PAL with an alpha channel as well as deep rasters, so they're quite heavy.

Can many small files benifit from high STR as long as they are sequentially stored on disk? If so, would it be necessary to defrag extremely often to keep them grouped? What am thinking, I guess that no single defragger is capable of organising numbered files? Does this mean that continuous movie files are the only way to go for maximum performance?

Then again perhaps compositing in general does not benefit much if at all from high STR, since even if continuous movie files are used, many of them will be retrieved in parallel. Therefore the drive heads would have to go all over the place to find all of the files and the streams would repeatedly be interrupted. Thus you would need one stripe-set per clip to keep the reads sequential, which would be ridiculous. Am I totally off here?

With this reasoning, maybe it doesn't matter if you store the clips as stills or movie files. The only thing that would matter then is low latency/seek/whatever. Please tell me what you think.

Thank you in advance!

03 March 2003, 01:05 AM
always use number stills with alpha channels

it makes it much easier to do compositing and after effects, plus you dont have to worry about codecs and things like that when you dont need to, then all you do is compress it to avi at the end and your done.

also, the seek time difference from multiple image files and one big movie is going to be pretty minimal as move drives now are damn fast anyway.

03 March 2003, 01:53 AM
Thank you DaForce. The main concern is whether to render the animation frames to multi-frame movie files (avi-style, but another format) to make better use of the raid 0 array. The reason being that raid 0 boosts sequential read/write operations but not arbitrary reads/writes. Thus it might be beneficial to use continuous files so that the disk array could read the entire clip as a sequential stream, without interruptions due to the need to move the drive heads to find the next part of the file.

As I wrote the original post I realized that since several clips will be loaded in parallel, there would still not be a sequential read operation. This is because several video streams would be composited together, so that each frame of the final composite would require at least one frame from each of the included clips. Thus it should not matter whether movie files or stills are used. At least in theory, that is.

As this reasoning is strictly theoretical, I would love to hear other people's thoughts and experiences on the subject.

CGTalk Moderation
01 January 2006, 04:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.