If I started several instances and let them render an image sequence simultaneously (with skip existing frames) it actually rendered quite faster if all of them were in single processing mode, compared to rendering the same sequence with just one instance of AE using multi processing mode.
This test computer I used is quite old, an eight core with only 8 gb ram. But increase in render speed was a little more than the double. The test project rendered was a relatively simple particle effect in Particular.
Here are the render times from various tests.
- one instance multiprocessing - 4:11
- two instances multiprocessing - 3:15
- one instance singleprocessing - 5:27
- two instances singleprocessing - 3:04
- three instances singleprocessing - 2:22
- four instances singleprocessing - 1:59
- six instances singleprocessing - 1:48
As you can see I get a 210% render boost with four or six simultaneous render instances compared to just one multiprocessing instance.
In the past there was a plugin from Gridiron that I believe automated this process in a good way. Unfortunately this plugin is no longer supported for CS6 and is no longer developed.
The process of manually getting multiple instances to render simultaneously is quite messy, at least on a Mac. It is something you want to do only once really. But it works perfectly on render slaves where you only set this up once using the AE Render Engine and watch folders. One benefit of this besides render speed is that you are allowed to run more instances of commercial plugins on each machine, as the plugin often doesn't need a unique serial number for each local instance you start.
And it seems multiple instances utilize CPU much more efficiently, as CPU usage stay more constant at the top when doing this.
Perhaps not all scenens will render faster using this method, but I guess one can build custom tools to exploit this more easily. Or why does not Adobe themselves fix their render system so that it uses system resources more efficiently?
Well, I'm curious what others might experience in terms of boosted render speeds using this approach.