I'm working on a 4K 32bps project (4096 x 2160) which has 12 particular instances emitting from multiple small masked regions of un-proxied .R3D footage. Each instance is set to emit 500,000 particles/sec with a max life of 3 seconds.
My machine is running OSX 10.6 with 2x 2.66 Intel Xeon (hyper-threading disabled) with 20GB ram. Rendering to an internal storage RAID.
I've precomped the particular layers and their emitters together and I tried rendering out a TIFF sequence (with alpha) to use as a pre-render, what I've encountered however is render times of up to 30 minutes per frame!
The next steps I'm going to take to try and speed up the renders:
Pre-render each particular instance on its own, then composite all of them back together
Try with hyperthreading enabled
Try with multiprocessing (though as my system doesn't have much ram I find this to normally be slower).
I've tried to simulate the same effect with fewer particles but havent been successful, which leads me to believe perhaps there isnt much I can do - though I'm hoping there is a magical 'render faster' switch that im missing ;P
How have others dealt with this kind of situation before?
Is this the norm with working with such high intensity comps?
Am i crazy for attempting this on a fairly average computer?
Have I done something completely stupid?
Should I start getting quotes to book out a render farm?