THat comes down to normal multithreading within a single process vs. Running multiple processes...
Thanks Mylenium! So to make sure I understand this correctly. The first four cores work on four separate processes meaning there are literally four iterations of AE open at the same time. While at the same time the other eight cores work just on the 1st iteration of AE and "help out" with whatever the 1st core is doing?
So if I have enough RAM to use all 12 cores then I would have 12 iterations of AE running at the same time and only one core working on each of the 12 iterations (no cores "helping" each other out)?
This is interesting then because I'd assume that if all 12 cores were working on one frame at a time each frame would finish drastically faster. However, rendering 12 frames at a time will take longer (per frame), but having them finish in bulk could make up for the difference compared to 12 cores working together.
It's obviously not a 1 to 1 ratio though; otherwise there would be no advantage to rendering multiple frames simultaneously. Which brings me to my last question:
What types of projects benefit from rendering multiple frames simultaneously, but slower vs single frames and a much higher rate? And the same vice versa? There has to be some way to get a rough idea without doing trial and error on every project.
There has to be some way to get a rough idea without doing trial and error on every project.
No, there isn't. Even minor differences across seemingly identical projects (another effect used etc.) can result in huge differences in processing behavior. It's completely unpredictable, even more so since even a theoretical ability to use MP rendering with all cores doesn't mean it will actually happen this way - limitations in CoDecs, effects, I/O bandwidth, audio processing and so on could thwart it. that and of course that there still would be multithreading in some of the processes as well as the master control process. E.g. the H.264 encoder is massively threaded and will always grab multiple cores on top of existing MP renders...
Thanks Mylenium! That's exactly what I was looking for. While that wasn't the answer I was hoping for, it did have all the information I was looking for. At least now I understand why some project perform better with different MP settings and how the cores get allocated while rendering which is the most important thing for me. Thanks again for sharing the technical knowledge!