Christoph Hormann wrote:

> I always found such artefacts quite noticeable,

I've read this many times but was never able to observe it in renders using
high-quality radiosity settings. In fact, I once made a "difference" picture to
test it by substracting the "non-stop" image from the "stop and resume" image
and I really had to raise the contrast level to see a few odd pixels here and
there, unnoticeable otherwise.

> I recently did a render at 2400x1350 which took 378 Mb instead of ~260 Mb
> at 1024x567 (same settings).

I've noticed this also, and it's quite fortunate that the relationship between
memory use and image size is not linear...

> Finally the split render would take longer because radiosity samples are
> taken several times, and render time IMO is often more of a problem than
> memory use.

I had tested this some time ago. The non-stop render took 68s and used 41 Mo
(160 x 120, no aa). I made three slices +er0.3 +er0.6 and final. Each slice took
14-15 Mo and the cumulated render time was 22+24+20 = 66s.
Also, if the image makes a large use of the swap file, the render time suffers a
lot. Actually, I started slicing renders because Pov kept running out of memory
on large images, and the slicing took care of the problem.

G.


--

**********************
http://www.oyonale.com
**********************
Graphic experiments
Pov-ray gallery

Post a reply to this message