Well, let's begin with the obvious:
in not too much time?
I'm afraid that doesn't make any sense on any level. Without knowing anything about the exact structure of a composition, any statement to that effect can only end up being useless nonsense. Sorry to be so blunt, but you can make AE slow to a crawl with a single effect, if only it is computationally intense enough and dealing with 4k and 8k panoramas and even larger source footage files isn't going to be a walk in the park one way or another. The file I/O alone could be a bottleneck. Likewise, most of your other considerations are moot at this point - AE barely uses GPU acceleration for rendering (though grapevine has it that this will be considerably pimped in the next release), sequentially writing to clip files has limitations on multithreading and so does using effects like Trapcode Particular that use linear progression to do their magic. Hence going crazy about Xeon systems or even dual processors is not really relevant - a lot of the cores will have nothing to do and you'll turn green with frustration on long renders. You simply need to understand how AE's rendering works in the first place and how it uses resources (or not). there isn't a single knob you can turn nor a single isolated hardware component that determines AE performance. Even maxing out RAM may not be relevant at all, since otehr factors will be more critical. You need to take a holistic approach.
Thank you for the information.
My question is more general, for a wide range of various effects like particles, glass, radial blur, ...
As I see, it is not easy to find the right hardware, especially when you don't want to waste resources or save at the wrong place. In this forum and other places I also found the information, that lesser but stronger CPU cores may be beneficial. Even though my CPU-monitor showed me, that AE and Media Encoder use all (virtual) 8 cores of my i7 Skylake, quad, up to 4,2 GHz, while rendering.
Anyway, I think I will go for a two GPU solution (GTX) and hope, that the Media Encoder and AE use them more and more.
What did you end up doing? It may help others trying to pull off the same workflow, so we are curious. Let us know!
Because Apple updated the MacPro (late2013), I acquired a model with a 6 core Xeon (3,6GHz), two AMD Fire 6 GB and 32 GB.
What I know definitely: 64 GB would have been better for 4K videos. Sometimes there comes a "soft" error message, that gives me the hint, that I need more RAM.
In some cases, I have the impression, that the Mac renders not very faster than my Windows PC (i7 4 core 4 GHz, GTX 1070, 32GB). Especially when the effects neither profit from GPU nor multi-threating. But still, it appears a bit faster and sometimes even clearly faster.
Some days ago I zoomed into a Mandelbrot fractal (4K). The Win10 PC had problems and got almost stuck after 7 hours and would have needed several more hours, so I aborted. The Mac Pro needed about 2 hours for a comparable fractal effect animation, even though a bit lesser computationally intensive. I monitored the CPU activity in both cases and all cores were on 100% almost all the time. The Xeon/Mac was way faster than the i7 Skylake / Win10.
Both computers do fine with the new GPU updated Fractal Noise, especially the Mac Pro.
When it comes to the Foam effect or particles, the difference is not that big.
However, I am not finished testing.
Anyway, the solution for me is still: pre-rendering of intensive elements of the video.
- 3D animations with C4D Lite/Cineware as a transparent Quick Time.
- rendering intense part of the video / backgrounds as a mp4 in a very high bitrate (e.g 80) and using it in the main video