Forum: Poser - OFFICIAL


Subject: PP12 Firefly Croaking and Telling Me To Do Stuff I Dunno How to Do

bloodsong opened this issue on Feb 11, 2012 ยท 24 posts


aRtBee posted Wed, 15 February 2012 at 3:31 AM

@Miss nancy - this adaptive bucket size is the Poser 7 way of handling resources, from Poser 8 / Poser Pro on it's replaced by the max bucket size methode.

Difference: the old way started with some bucket size, and grew it with a rate (the lower the threshold the faster) until it met resource limits. The new way start at the given max (default: 32) and shrinks when it meets resource limits (acc to the manual).

What happens: each thread gets a bucket to render, each one has size (say 32x32 pixels) plus some overhead (say 8 pixel border, I don't know the real number). In this case a (40x40/(32x32) = 156% overhead. When the thread is finished, it gets a new bucket. Poser cannot render individual buckets over the network (Bryce and MojoWorld can). When you double the bucket size, the border stays intact. So the memory required to the thread quadruples but overhead reduces to (72x72)/(64x64) = 126%. More memory, less CPU time.

notes:

(1) when you press Cancel, finished threads will not get a new bucket, but the running ones are not stopped. So you have to wait longer for a response.

(2) if you've got many threads available, large buckets will create a situation where just a few are runnning at the end of the image while all others remain idle. Example: a 500x500 image at a 64-bucket size dicides into 8x8 = 64 buckets, so on a 12-thread machine the last porting will have 4 active and 8 idle threads. Only when the last thread is done, the image is finished. Hence the question is if larger bucket sizes really give an earlier result. In this example: render time will be reduced to 88% instead of the expected 126/156=82%.

(3) less threads running will have less simultaneous buckets worked on. A bit less memory use, far less CPU power, so far longer render time. Your image, your call.

(4) smaller buckets (16) are adviced for scenes with fine detail, like a facial portrait with dynamic / complex hair or much displacement mapping. Larger buckets (64) run fine on scenes with buildings, distant figures and props. Generally, the default (32) fits both.

- - - - - 

Usually I'm wrong. But to be effective and efficient, I don't need to be correct or accurate.

visit www.aRtBeeWeb.nl (works) or Missing Manuals (tutorials & reviews) - both need an update though