Anim8or Community
General Category => General Anim8or Forum => Topic started by: Raxx on December 02, 2012, 11:19:25 pm
-
Hey folks. I figured I'd post this for those who haven't discovered it yet. I rarely render animations so I haven't bothered looking into speeding up render times until now, but after observing Anim8or's behavior with multithreaded rendering, I discovered two things:
- One Anim8or rendering instance takes up a maximum of one cpu core capacity
- You get the same render speed even if you have as many renders going on the same time as your number of cpu cores
For example, I have an 8-core cpu (technically 4 cores with 2 threads each). One rendering instance takes up 1/8th of my total CPU capacity (or 12.5%). This means I can have 8 Anim8or programs running and rendering at the same time without losing speed for each rendering instance, assuming I don't run out of RAM (nowadays that's not a problem). Doing this on a single animation will decrease my total rendering time by a factor of 8. It still takes hours and hours for a simple scene like the attached image, but it's a good deal better than without.
(I'm rendering at 75% capacity for the sake of multitasking)
-
I used anim8or like that for a long time, since at my job I sometimes need to render multiple files in the same time :D. Good to mention this on here. :D
-
I've done that as well, but usually I need to utilize all of my cores on one image, so I just go to Carrara for that
-
Getting dual core next year, interesting post, thank you
-
One question Raxx: do you know if switching off multi-threading (ie. just using the 4 cores) would effectively double a single render instance running on a single core? I have been planning to test this since I often need to render single frames or short frame sets as animation tests, and running multiple rendering instances isn't practical in this case. With multi-threading I am currently using 1/8th of the CPU, as you say, but with it switched off it could potentially use 1/4 for stills.
-
With multi-threading turned off, it seems it still gets split across the cores in the same percentage. However I had found that not only would it not visually update the render until it was done, but it had a good chance of crashing as well.
I should also note that I'm using Windows 7. Such behavior may be different in older operating systems.
Do you guys remember that tool someone made to increase rendering times of single frames by splitting it in parts by covering a portion of the camera's view? Not sure how effective it'd be in ART, but it did work for the scanline rendering, I believe. Another idea is to render a still frame normally in one instance, and then flip the camera upside down and render it in another instance, and then halfway print+screen and splice together...though that'd only work for images less than the screen resolution.
-
I'm am TOTALLY fina use this. Got a big animation proj coming up...