I have a Comp, which admittidely is a render nightmare, with 64 instances of a the same video each 1 frame ahead of the next. So it's very RAM hungry and something of a balancing act between the number of CPUs and how much RAM each is allowed (but generally allowing 3GB per process and 1 less CPU than suggested)
The thing that is killing me though is the RAM Preview is going almost 3x faster than a normal render. Even when the normal render destination is an SSD and compression is None. otherwise exact same settings i.e. 100%, best quality.
The really interesting part is that when doing a MP RAM preview the CPU usage (as viewed in activity monitor) is a solid wall of green but when I do a standard MP render the CPU useage is fluctuating wildly. I've tried various dififerent MP options (all maintaining a safe RAM overhead for the sysytem) with very little difference in performance.
This is driving me crazy.
MacPRO 2x 2.26 QuadCore, OSX10.8.3, 32GB RAM, Proavio RAID for source, internal SSD for destination.
Well, clip formats need to store there frames consecutively, don't day? You can't expect MP rendering to defeat that. If you want to max out performance, use image sequences, otehrwise you'll have to live with this limitation of each instance having to wait for the otehrs to have written the frames in proper order.
Right, so does that mean the Ram Preview is writing frames non-consecutively and that's why it's so much faster ? I still don't understand why the CPU useage would be so different.
I wish I could use image sequences but these are 1-2 hour videos that are being rendered.
One to two hour movies produced entirely in AE may not be the best use of the software. It's hard to tell with knowing more.
Image sequences can be as long as you like. Most of the renders that visual effects houses produce when making a feature film are image sequences. They are just more efficient. A jpeg or png QuickTime is just an image sequence in a .mov wrapper. If I were attempting to render a two hour project in AE I don't think I would even consider rendering to a format that may break if the render fails at some point. With an image sequence you can pick it up to fix just one frame if you need to.
As Mylenium said there are a lot of limitations with system resources that come into play when you are rendering to a video codec. Some do not support multi processing at all, some have limited support. There's no reason yet to expect that rendering to a video codec will be as efficient as rendering to an image sequence.
Jeff, sorry to hijack the thread, but I have a question...
If the JPEG Quicktime file is just a series of JPEG images in a quicktime wrapper, why choose one over the other?
Are they technically exactly the same?