Hello,
I am currently using Vray 1.5 with Rhino 5. I am a bit perplexed as to why I can easily render an image of 1000 x 750 in about half an hour. If I keep the same setting but increase the size to 3000 x 2250 it will take 20 hours. Logic would say that a rendering 3 times the size of one that takes half and hour would take 3 times as long to render (1.5 hours).
Does anyone know why the render time is increasing exponentially and if there is something I can tweak to correct this?
Thank you in advance.
Well it’s not 3 x bigger, it’s 9x (3x as many pixels across x 3x pixels down). But that still would only make a 3 hr rendering if its a linear increase from 20 mins. I have also been seeing this problem but only on certain files and haven’t figured it out. What settings are you using? Both IRMapand LC can be affected by the output size, and the image sampler definitely is. Which pass is taking longest? On my problem file LC went from a few minutes to many many hours just by 2x output size (each direction, so 4x) but I had LC set to screen size so I don’t know why it should take so much longer?
Hello Fooprobe,
Thank you for the reply. I understand what you are saying but am hoping that there might be a way that I can still speed things up. I am using a new computer with 3.4 GHZ Quad-core Intel Core i7 with 32GB of Ram and am curious why this isn’t enough computer to render something large.
I should also say that I am rendering from both a saved vrlmap file for both the Irradiance Map and Light Cache because I have heard this speeds things up. I also tried rendering to a vrimg file because I heard this was better for large images (it did not seem to speed things up at all).
Here are my settings for the rendering:
Small Output (1000 x 772 pixels)
Large Output (3300 x 2549 pixels)
Color Mapping:
Exponential
Dark Multiplier 1.0
Bright Multiplpier 1.0
Gamma 2.2
Input Gamma 2.2
Indirect Illumination (GI):
On
Post-Processing
Saturation 0.8
Contrast Base 0.5
Contrast 1.0
Primary Bounces Multiplier 1.0 Irradiance Map
Seconday Bounces Mulitplier 1.0 Light Cache
GI Caustics (Only Refractive Checked)
Ambient Occlusion On (Checked)
Amount 0.8
Subdivs 8
Radius 10.0
Irradiance Map:
Basic Parameters -5
Max Rate -2
HSph. subdivs 50
Interp. samples 20
Clr thresh 0.4
Nrm Thresh 0.3
Dist. thresh 0.1
Interp. frames 2
Light Cache:
Calculation Parameters
Subdivs 1000
Sample Size .02
Scale Screen
Number of passes 8
Depth 100
Store Direct Light (Checked)
Show calc phase (Checked)
Adaptive Tracing (Unchecked)
Use directions only (Unchecked)
Use directions only
Min paths per sample 16
If you’re using saved lighting then your settings (mostly) dont matter. the bulk of the render time is in the image sampler (except loading the lighting maps, but that should be the same for both). So research optimizing this - it’s going to depend on your specific scene. Watch out for reflective and refractive materials, and for anything blurry. Try to reduce the number of point lights if possible.
it’s not your hardware. It’s really easy to set up renderings that take 20 hours on even the fastest machine.
you also may find you get better contrast if you reduce your LC multiplier to 0.7 in indirect illum, but this won’t affect render time
IIRC rendering to vrimg isnt a time saver, it’s just going to save ram by skipping the frame buffer, but that doesn’t seem to be your problem with 32GB - unless you somehow don’t have a 64bit system 
Hi again Fooprobe,
I think you’re right that it is reflective/refractive materials that are slowing things down. I am using Architectural Glass as a material and the render seems to slow down when it gets to those parts. Do you have any recommended setting for glass without scarifying too much quality? Thanks.
Everything is a trade-off, there’s no free lunch. anything you do to reduce time will also reduce quality, unless you have things cranked way too high. If using blurry reflection/refraction, the more blur you have the slower it will render. You can try to reduce the subdivs but it will look grainy. You might also look at your reflect/refract max depth and make sure it’s not too high (you might try it as low as 2) - if the glass goes black, increase it
You might also make sure that any material with refraction is on a solid - when on single surfaces the refraction is exaggerated and I’ve seen that slow things down too.
I don’t understand the strange render time increase, also I never have seen this effect. Could be helpful to see the rendering or get the scene to test it.
General the LC pass is independent from the output size, also at high res renderings the IM max rate can be set lower, so that my high res renderings was faster than I expected from the small test renderings.
Some general hints: it’s strong recommended to set the max refl/level low as possible. I rendered a lot of interiors (where no mirror show reflective metals) with max refl/refrac depth 1. I used glass without refraction, only reflection + black diffuse color + diffuse transparency color for tinting the glass. If dark edges of glass sheets was needed, than I used a dark version of the material for the edges. Some times I used a refractive glass for the sheet edges only. If I have right seen, rectangular lights are faster calculated than spot/points, since there “store with IM” can be enabled. Important for scenes with a lot of lights.
Frozen glass sheet can be rendered with depth 1 too, if the IOR is lowered to a low count like 1.05 and a light glossiness only. I needed refl/refrac depth 2 only if metals was placed behind frozen glass, but this increased the render time a lot. Frozen glass can be much faster rendered with enabled interpolation.
I’m comforted to know you think this is weird too, micha. But I definitely have it on a file I was working on too. Way too big to upload. Ill try to see if its a reflection thing, I can’t remember
I am using Vray 1.5 with Rhino 5. I don’t think that you can control the reflect/refact max depth the same way as you can in 3d Studio Max.
How can be controlled the reflect/refact max depth at 3dsmax?
maybe not the “same” way, i’m not so familiar with Vrayfor3DSMax. but in VfR, in global switches there’s an option for the refraction/reflection max depth.
Hi Fooprobe,
I tried the “Fake Glass” Micha suggested, I did find the refraction/reflection max depth in the Global Switches and played with that, and still I have experienced no difference in render speed.
I sent the scene to Micha via dropbox to see if he/she can find anything else that may be making it render for so long.
I could send it to you as well if you think you could take a look at it.
Thanks,
Tim
Hello Timmy11,
Could you please send the scene to support@chaosgroup.com
We will try to help you with this issue.
Thank you very much in advance.
Hi Svetlovar, if you find the solution, it would be really great if you could post it here! Thanks!
Hello Tim
Unfortunately we are not able to reproduce the issue in our environment.
Here are the results we got:
500x386 = 3min 14s
1000x772 = 10min 38s
2000x1545 = 37min 4s
3000x2318 =1h 18min 2s
Some of the textures are missing - from what I can see these are the textures on the trees around the building and the cut-wood under the roof on the right side of the building. The other difference is that my images have less contrast - probably some parameters are changed between your renders - and the scene you sent us.
I have noticed that 3k resolution consumes a lot of ram more than 16GB - don’t know how much exactly because my machine has 16Gb and when the render comes to 30% it started to swap memory - which is something that affects render-time. In your post you mentioned that you have 32GB of ram - so this shouldn’t be an issue.
The other difference between your and my rendering except the textures is different PC configurations - and probably Vray version. In my tests I used Vray Nighly Build version 22829 - and you have probably used the last official Vray version.
Could you please inform us what Vray version are you using.
Also - do you have any other machines available for test renders ? It is better if they have at least 16GB of RAM. If you have such a machine - please run one more test on it - and let me know if the result is the same or it is different.
I found no time to test the scene, but I think the problem could be related to the displacement of the grass. Svetlozar, maybe in your test the displacement map was missing and you tested the scene without displacement.
General for displacement I found a speed turbo trick. Read here:
http://www.chaosgroup.com/forums/vbulletin/showthread.php?63305-Displacement-for-arch-viz-a-speed-up-observation-and-more
It’s a pity that this trick isn’t implemented as standard option.
I started a rendering - the LC pass is horrible slow, caused by the displacement. If I set the global displacement amount 0, the GI is quite fast calculated. Saved to the hard disk as cache, set to read and displacement back to amount 1 and the rendering run much better.
At the global displacement option are some critical options, careful to handle. I use displacement quite seldom, so I’m not sure about the options. But “view dependent” and 256 subdivs sounds “dangerous” for high res renderings.
Thanks Micha for looking at those but in the images Svetlozar sent me the displacement seems to be the same (the materials with displacement, grass and concrete, look the same in his materials as they do in mine). My computer actually had no problem rendering the trees and grass in the image, it seemed to be when it tried to calculate the glazing that it started to freeze.
In order to get to the bottom of this I will need to know what Svetlozar’s computer settings are in order to match mine to theirs. Also if they are using a different release of Vray 1.5 then me. Those are the only variables that could possibly be causing me problems. It does seem like a bug as my computer renders about the same for the 2 smaller sizes (500x386 = 3min 14s and 1000x772 = 10min 38s) but is not even close to the times mentioned for the 2 larger images.