Hi all,
just wanted to share my first BF/LC render 
took 35 hours accross 3 PCs at 3.5K rendersize
IM/LC version took 6x less :)))
Best Regards
unfortunatelly forum keeps rezizing my images all the time…
here is a 1500px version
http://www.pixelbox.cz/fileadmin/Preview/meland%20final\_POST\_1500px.jpg
Martin can you show IM/LC version ?
Yep will do when i get to office in the morning… Will show some comparison details
Looks great Martin! Time seems long to me though. What were your DMC and BF/LC settings, out of curiosity?
b
Looking very nice. Is reflection Gı also enabled?( I see some caustics like reflections under the table.) Maybe that is taking long time.
thanks guys!
the image is missing the most important object which we are still modelling…but thats another story
LUKX:
here is the IM / LC with no PS
IM/LC settings were
IM (-5/-3, HS 80, IS 20)
LC (1500, 0.02, prefilter 100, retrace threshold 1)
DMC AA 1/8, clr thres 0.002
and the BF with no PS
i can send you full 3500px images if you PM me your mail
BF settings were:
DMC 1/16, clr thres 0.002
BF subdivs at 70
LC 1500, 0.02, prefilter 100, retrace threshold 1
Pixela:
yes i ve noticed now it was enabled :)…could add to the rendertime quite a deal
Brett:
the BF etting might have been an overkill but as i was doing it for the first time, i went for quite high settings, which i mainly copied from this post:
http://www.chaosgroup.com/forums/vbulletin/showthread.php?54547-About-noise-in-general
please read the BF settings above
ALSO i had to raise a few individual material subdivs from default 8 to 24 (wood, floor, some metals)
BF times could probably be tweaked down quite a lot but i did not have time to experiment with it yet
anyway, i absolutely love the way it handles the shadows…thats unbeatable!!!
I will try to persuade my colleague to get me links to the hires images so i can actually post them here (links at least) :)…as soon as he gets here
BTW:
why does the forum resize the images so much??? i think that only mine get this small :))) am i doing anything wrong?
All the best
here are the 3500px links if anyone is interested comparing the qualities:
BF + post
http://www.pixelbox.cz/fileadmin/Preview/meland%20final\_POST.jpg
BF no post
http://www.pixelbox.cz/fileadmin/Preview/meland%20final\_BF.jpg
IM + LC no post
http://www.pixelbox.cz/fileadmin/Preview/meland%20final\_LC.jpg
Nice image. I have to say, the IM version is pretty good. Doesn’t show any of the typical IM shortcomings. Render time?
Edit: do you have the Noise threshold value for both renders?
thanks (and thanks for your great models :)))))
6 hours 30 minutes
DR 3 comps
1 Sandybridge @ 3.4Ghz, + 2x i7 at 3.4Ghz
think if i enable DE and widen the AA scizzors more, i could get closer to BF
by looking at your other post about noise in general, i also have to admit vray uses muc more RAM than it used to do :-/
it’s pretty good, yes. you can still spot the obvious areas where irr maps fall short though: contact shadows and generally very fine details around the books, the carpet and the side table. but at that resolution, once you scale it down, the differences are almost invisible. it’s a very good image, regardless of the gi method
I would definitely experiment with much lower BF bounces. I typically run at around 15 and my noise threshold is rarely much below .01 - usually around .008 For noise control I prefer to crank up the material settings for problematic materials - but I tend to start around 24 and go up from there. For a glossy leather like that chair I’d probably set it to 40-50 and see how it went. My feeling is that you can drop your BF render times considerably, but hard to say where the noise might become unacceptable for your taste.
The IM/LC one looks pretty good to me - I’d be happy with that version, and much happier with the render times
Nice work in any case.
b
but does it do anything when using BF/LC method, cranking up material settings (subdivs) ? \i don’t see much difference, only bringing down noise threshold in dmc sampler is making glossy reflections smoother.
Brett:
By saying BF bounces do you mean subdivs?
I have similar approach with individual mats subdivs when renderin with IM/LC.
The only reason (apart from serious urge to finally try BF) i went for such over the top settings was that i had free comps for rendering over the weekend :-)))
Thanks for the kind words guys!
I think it depends a lot on how high your DMC sampler settings are, more than the BF/LC settings - that is my understanding anyway (which is limited). It probably depends a lot on where the noise is coming from too, but I’ve sort of settled on that range of value for my materials so I don’t have to mess around with as much testing. Seems to work for me, but it may just be my own workflow/type of images I do. Glossy refractions are another story though - they need to be very high (100+) and are crazy slow to render
I actually had a Maxwell image cook faster on my last glossy refractive material images 
/b
Yes, sub-divisions. Sorry about that
i feel like 1/16 for DMC AA could be a nice starting point for me now that i ve started dig in the BF/LC more…
how about you Brett?
whats your experience with glosier scenes??
I have usually defaulted to 1/40 actually, but lately I’ve been using 2/20 to avoid some problems that popped up on a job a while ago. I forget the exact circumstance, but Vlado suggested raising the min value to 2 and it worked. I did not really note any hit in render times so I left it that way since
If all looks good then I might experiment with lowering the max value, but I really can’t say I ever noticed a huge difference in render times so unless the render seems unusually slow I don’t bother. Perhaps on smaller renders with many frames (animation etc) it makes a difference, but for hi-res stills I didn’t notice much loss/gain between 20 and 40 in either quality or render times. I have not tested it exhaustively though…
Anyway, I very rarely need to do more than tweak the noise threshold for DMC - generally my GI settings stay the same on almost all jobs. Been experimenting a bit with IM/LC lately, but it honestly seems like more trouble than it’s worth. I wasn’t getting a huge difference in render time, but probably that is my lack of experience with it.
/b
/b
Let’s not forget that we might also smudge the noise a bit with a bit of AREA 2.0 Antialiasing filter
I never use AA filtering at all I actually like a bit of noise in my renders. I’m often adding them to photos, or using them to replace photos, and that bit of “tooth” gives them a more realistic quality. Even with complex materials there it often helps to add a bit of surface complexity that adds realism.
Just my POV.
/b