Hi AE community.
I have a pretty solid computer but am very disappointed with how slow frames update in After Effects. I am quite familiar with how to set up AE for RAM/CPU distribution, but maybe I've clicked something that is causing this issue.
My system has 24 cores and 64 gb of RAM.
I allow AE to use 8 of the 24 cores and give each one 4GB of RAM
Also have CUDA enabled for a gtx580 3GB.
OS is Mac 10.8.5 with AE Creative Cloud, fully updated along with the NVidia drivers.
I'm attaching an image (below) that shows 12 of my 24 cores jumping to action during a frame render (strange b/c I told AE to only use 8, but I am not complaining... the more the merrier), then one single core finishes the render very slowly.
Another point worth mentioning is that I am rendering a frame for a map animation that uses a 20K jpg that is 10.7MB in size. There are a few images like this overlaying one another. The 10.7MB size is actually quite small given the resolution, but perhaps it is an issue... although I don't know how that relates to AE simply not using the immense power at its disposal.
I am using the CLASSIC 3D renderer. The Raytraced 3D renderer is blazing fast, but I can't use my various overlays that are needed, so its not a viable option for me right now.
I'm still stuck wondering why all my CPU power is going to waste within the program. Perhaps its worth mentioning that when I do a final render to file (especially using the Background Renderer script) all 24 cores are utilized and frames are rendered so much faster than when I am still working in my comp.
Quick info on the image:
#1 = what happened the first 10 seconds after moving to a new frame in the timeline
#2 = another 30 seconds of render on one core only
Thanks for the reply. No luck when turning off disk cache and gpu hardware acceleration was not on.
I'm still very stumped as to why AE refuses to use all cores (even just more than 1) when working in-comp, but will use all when rendering a final version... The disparity in speed is enourmous and I don't see how AE justifies the difference in hardware utilization.
Mylenium / anyone else,
#1: I've run a number of tests. I swapped my 20k jpgs for psd files after someone had mentioned it may be the decompressing of 20k worth of image. The render time was pretty much the same.
#2: I found the real culprit to be Trapcode Particluar. I have Particular 2.2 running and it was spitting out a lot of particles. I have found that when I turn off that layer, everything renders very quickly, even my 20k images in less than 3 seconds. I then turned Particular on again and reduced the number of particles by more than half. My render time dropped from 30-40 seconds, down to 12 seconds. That is great... still not as fast as I would like though because AE/Trapcode still only uses one core to calculate and render.
Now that this has been narrowed down to a Trapcode issue, can anyone explain why it is only using one core?
> Now that this has been narrowed down to a Trapcode issue, can anyone explain why it is only using one core?
There's a lot of work that goes into writing multithreaded software. You'd need to ask Trapcode folks why Particular isn't multithreaded in this case.
I am aware that MP is complex. When I said it has been narrowed down to a Trapcode issue, I meant that in this case, Trapcode is my biggest issue for a slow response, however, AE still underpowers itself even when I am working in comps that have no Trapcode or third party plugins involved.
AE is just inefficient and I don't know why.
I am attaching an image of a simple render - two 20k jpgs overlayed on one another. That is it. The render takes about 5 seconds, and you can see the CPU activity that takes place is primarily on 1 core out of 24. Even though I have reduced the number of cores AE can use down to 8, with 4GB each - only using 1 is very ineffient and I wonder why AE CC is not taking advantage of my computer. That 5 seconds could surely drop to 1 if those 8 cores I gave AE were utilized.
To this end, when I render using the Background Renderer Script as seen here: http://aescripts.com/bg-renderer/ the utilization of my computer is far beyond what AE will natively do. This wouldn't be a problem if the BG Renderer script also sped up the in-comp render speed, but obviously it doesnt and I am forced to wait on AE to do it in its own sweet time.
Thanks for any further insight you have.
There's no further insight to give, other than perhps to tell you that we currently have a large fraction of our team working on making all of After Effects behave better (i.e., faster) in this area.
That is actually GREAT to hear. Outside of telling me that I just needed to change a setting and 'viola', this is the best news you could have given me. Give my regards to 'the binary guys' and I will look forward to their update. Any idea on a timeline for any major breakthroughs? Weeks/months?
I want to add a final note in here. I did some additional testing this morning and was able to find a solid sweet spot for AE based MP rendering. It was a 1920 comp with a large jpg that had AE effects, Ball Action and Add Grain applied. Without MP the render was about 27 minutes, with MP is was just over 2. The computer utilized about 60% of itself for this render, which is better than normal. When I used the Background Renderer that I mentioned above, in this case, it was slower, coming in around 5 minutes. The BG Renderer seems just as hit or miss as AE itself. Largely depends on the comp.