Tweet
Small buckets are bad for ir map calculation. Try to use buckets of 50px wide and then min/max to -8-1 just as a test. It will do pass 1 and then skips pas 2/3/4 because buckets are too small. Imo, 64 is the best all round value, and with high resolutions (2400px and more), 128 is better.
-
could be wrong
I could be wrong on this one, but it looks like Jujubee is using the single frame methond for his ir map. Couldn't he calculate the irmap at a lower resolution, say 1445x935 and use that for his final image? It won't take care of the noise, but once he's figured out those settings at a lower resolution, it would speed up the final rendering.Comment
-
I did some tests before, and My render times were faster using a larger buckets. 128 works better than 64, atleast in my experience.Originally posted by flipsideComment
-
Correct me if im wrong (probabaly am) but when you increase the resolution cant you decrease the IRmap min/max rates and still get equal results to that of a smaller res render with higher min/max rates.
Like couldnt juju drop his to -5/-3 and still get a decent looking render?
Comment
-
I believe you are correct, at least that's what I have read. It seems to work for me.Originally posted by DaForceChris
The RevitlutionComment
-
Is there any formula/ratio list to this size/res vs. min/max rates???Like couldnt juju drop his to -5/-3 and still get a decent looking render?Comment
-
I think you you can lower by 1 each time the resolutions doubles...mmm saw it somewhere let me look..... ahhh ha
hehehe its in your own thread!!! 4th post from the bottom by s0real
http://www.chaosgroup.com/forum/phpB...lution+min+max Comment
-
that's bad when the answer lies within your own post. I'm going to go find a rock to hide under now:
Flipside mentioned -3, 0 at 800x600 in that post but it was mentioned here that 0 doesnt allow for undersampling.Originally posted by s0realComment
-
Well it does and it doesn't
If the samples were set to 0,0 then there would be no undersampling. At -3,0 some parts of the image will be undersampled at -3 ( pixels/4 samples), -2 , -1 and then 0 ( 1 sample per pixel) depending on the threshold values. Atleast that is how I understand it...Comment
-
where's that rock again?Comment
-
Guys, IMHO with this size of a rendering (2900x1900) set IRR map to "very low" preset or even faster like ~-5;-3.
It makes no sense to add such a fine detail to the big print sized 11x17. Lots of those IRR map tiny details get lost in print and under your textures
anyways.
I recommend IRR settings:
- "very low" hsph 60, interp 35 (to smoothen things out - cause you`re using low settings).
Then I would fiddle with QMC:
- starting from adapt 0,9, noise 0,005 and going up when needed. I would leave the min samples at their default here or even strike it to 10.
Remember, bigger renderings do not neccessarily need higher settings! (IMHO - basing on my experience)
Cheers!Comment
-
It's pretty easy really. -3/0 means that:
- in pass 1 (-3): IR map is calculated for the image at /8 its size (2*2*2). So for an 800*600 image the first pass is calculated for 100*75.
- pass 2 (-2): Info of pass 1 is used, and refined for the image at /4 size
- pass 3 (-1): Info of pass 2 is used, and refined for the image at /2 size
- pass 4 (0): Info of pass 3 isused, and refined for the image at full size
So the undersampling lies in the first 3 passes. The last pass is not undersampled but since it uses the info of the previous passes it doesn't need to refine the GI in all areas. (vray knows where the 'difficult' GI areas are, and only if they don't meet up to the thresholds like clr and dist, there will be taken more samples)
So the undersampling is simply used to find out where the GI needs lots of attention and where not.
This is also why the first pass can take a long time compared to the second for example. In the first pass, there is no info of a previous one available.
This also explains why too small bucket size is not good for IR map caclulations. Vray computes the IR map for each bucket, for example of size 50*50. First pass with min/max at -5/2: 50/2/2/2/2/2 ---> ir will be computed for a 1.5625*1.5625 bucket size... Not very usefull. So for high res renders where you usually use lower min/max rates, use bigger buckets!! Also each bucket needs to be adjusted to its neighbouring one to make sure you don't see sudden shifts in GI light, the less buckets you have, the less adaptation will be needed.
A similar post of me was added to vray.info a long time ago:
http://www.vray.info/topics/t0101.aspComment
Comment