View Full Version : Is my render time acceptable for this image?
02-21-2005, 05:02 AM
This image took 4hrs 4 min to render using Mental Ray with Gi and FG on a p4 3.0 ghz (overclocked to 3.45 ghz) and 512 of ram. The scene has 4 photon emitting area lights. Should my scene render faster than this, or is this an accepted time for a scene such as mine. here's the image....
(suggestions for ways to improve render time too)
02-21-2005, 05:37 AM
Heh, 4 hours for that resolution is WAY too much. It should render in about 10 minutes tops.
you waited 4 hours for this poor image... hmm... what settings have you used for antialiasing?
02-21-2005, 09:14 AM
Yes, it seems the rendering time it too long. Maybe you're using a high sampling in your area lights or in the whole global setting.
Just go slowly in your lighting, Maybe you'll need to use shadow maps for some lights? or get rid of the shadows completly.
02-21-2005, 09:16 AM
'you waited 4 hours for this poor image.'
wow, asshole.. what's wrong with it? I think it looks pretty damn realistic.
i donīt think i looks realistic, but i like it anyway ...
lol, asshole :)
I think the modeling is good, but the lighting it's poor, at least this is my opinion. You should make only DGS shaders for this grey room and the lights should have mr_physical attached to them (so you can just copy the V on the light color to the photon intensity). Also, if you don't have maya 6.5, to get realistic results you must use both GI and FG.
click here (http://www.impresszio.hu/szabolcs/MentalRay/MentalRaySampling.htm) to learn about Adaptive Sampling
02-21-2005, 10:51 AM
won't using dgs shaders and physical lights make my render time go up even more though? I hadn't tried that yet just because I assumed it would. I'll post my render settings and light setting later whe I get a chance so maybe there's somthing I have set WAY wrong. Also, as asked earlier, my sampling is 0,2 with a contrast threshold of .050 for all RGBA and a Mitchell Filter at the default 4.0 setting. Also, what do you mean by saying that if I don't have 6.5 I need to us FG as well to create realistic lighting? Don't you have to use that in both situations to get the most realistic lighting? Anyways, i'll post later. Thanks so far everybody. Oh and while i'm at it, i'll post my reference image for this piece so you can see what i'm going for as well..... http://img202.exs.cx/img202/1294/stairwellreferencephoto9gl.th.jpg (http://img202.exs.cx/my.php?loc=img202&image=stairwellreferencephoto9gl.jpg)
02-21-2005, 11:38 AM
Perhaps this will help a bit:
As for rendering time I think it is TOO much .
I guess one of the main reasons is the usage of 4 Area lights that are expensive to render instead you can switch them into spots( Emmision from area lights is much more expensive than other lights )
2nd thing I guess you are using either high detailed Raytrace shadows or very high res Dmap shadows with high filtering which in this case with all these 4 lights I suggest to bake the shadows .
Over all I suggest to minimize the usage of all these lights and use an environment light ( as you see in the refrence the light is an ambient daylight ) .
And I guess the rays used for FG should be very high.
All together I suggest to get rid of FG and GI since I beleive for such a scene you just need a dirtmap.
At this time you must be around 20 minutes per frame Maximum.
At the end Send us some Wire ( I beleive some objects and all those bevels are high poly ) and mental ray render summery to see what is the most time consuming part.
Hope it works
( My collegue is sitting here is saying what you are trying to tell him is becoming a tutorial )
Any way I should admit I liked the lighting.
02-21-2005, 01:54 PM
Youre not using point lights with depth map shadows are you?
I would fake all the radiosity based on that image. I mean what you see there can easily be done with using real lights as the bounce lights.
02-21-2005, 02:41 PM
why are you using FG here? I know its a neat effect, but you really have no use for it!Use proper ligthning and only GI with tuned in accuracy and intensitivity and you should get an even better render after a bit of playing
(i think the shadows edges are a bit to sharp, atleast if the aim is realism)
and a greatly decreased rendertime!
02-21-2005, 07:51 PM
welllll......looks like i'm getting a unionimous "you're not doing somthing right" lolI'll take the suggestions you guys gave me and try going a different route with this render. I had originally used straight GI, but I just wasn't able to get th e accuracy of the lighting for some reason?? i dunno....this is all a learning experience for me as I have never used MR in an actual scene before. I have done many many many tests and stuff only. Great help you guys have been. Hopefully i'll post somthing much faster in the near future lol
02-21-2005, 09:52 PM
Part of it might also be that all your objects are NURBS and are really highly tessellated?
02-21-2005, 09:59 PM
welllll......looks like i'm getting a unionimous "you're not doing somthing right" lolI'll take the suggestions you guys gave me and try going a different route with this render.
You could always post the scene here, maybe someone download it then try to setup the light things !!
02-22-2005, 04:12 AM
As far as the nurbs comment......It is a poly object and is only 33,000 polys. I'm noticing that the photon emission goes smooth and fast. I'm testing right now with just 1 area, physical light emitting 1 million photons and the calculation only takes about 10 seconds.
"RCGI 0.2 info : optimizing access to globillum Photon Map with 1000263 photons
RC 0.2 progr: computing camera volumes"
once it hits these, it really slows down, and then the "rendering" section is slow as well. I'll post my render globals and the scene once i get some time. Thaks so far guys!
02-22-2005, 07:47 AM
Think i found out my problem of why it is taking so long. When I was using just GI, i had each light emitting 1-2 million photons, which made the image very smooth and still rendered quite quick....(lik 20 min) and I had kept that same setting when I started using FG. Upon reading on this forum I found out that when using FG you can turn down the amount of photons tremendously, which I hadn't been doing. I'm doing some tests with a single light right now and it looks like a single light will render at around 10 minutes which in theory will make the 4 lights go around 40 min. Which is completely doable for me as this is going to just be a still shot. does that sound right now? New question too....if I lower the quality of the fg some, so that it is slightly blotchy, but still nice looking, will this be unnoticable when I finally texture the scene?
02-22-2005, 07:53 AM
textures will definetly hide some of blotchines, but i think you will not know until you won't try it.
02-22-2005, 09:09 AM
You mentioned you got just 512 MB of RAM. To me this sounds a bit less, when taking into account that Windows XP alone needs at least 200MB. Rendering times might increase dramatically when virtual memory is being used.
Just a guess ...
02-22-2005, 09:23 AM
I'm buying another gig of ram a the end of this week, so I guess I will find out about that soon enough :) I guess I didn't get my time down as low as I thought. That time I posted was with the sampling at 0,1....which isn't really good enough...when I uped it to 0,2 it went up to 40 mionutes :( anyways, here's my scene if anyone wants to take a look at it. I have a bookmark set for the renderview, and I currently have all lights hidden but 1. errr...actually, how to I post a file on here? I have only ever posted images lol.
02-22-2005, 12:11 PM
Dude, just get the Turtle Renderer from Illuminate Labs. It looks better and its WAY faster.
Its just amazing. I don't think the trial is availible anymore. That sucks. Try searching the net for a copy... http://www.illuminatelabs.com/
02-22-2005, 12:32 PM
lol Phil_Boy, thats one strange statement, but whatever..
Using tons of photons increases RAM usage dramatically, so this could be a reason for slowdown indeed. Try something like 100.000 to 200.000 photons per light, turn down the GI accuracy to a low value (like 128 or even lower) and turn on finalgather.
Then try to break down the area light sampling to an acceptable (speed vs. quality) value - maybe 4/4 or even try 1/1 and see what happens. Start with a very low value and then move slowly forward.
If you encounter finalgather blotches try to increase the finalgather filter to 1 or 2. Only increase this if you certainly need it because it will remove contrast of your finalgather solution.
02-22-2006, 01:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.