need help on render times--any tips??

[ddho1981]

I am posting two images ... the one looking towards the kitchen took about 9 hours to render at 2560x1920, but the one looking out to the window wall from the kitchen took over 20 hours! I know reflection is a huge factor -- any tip on how to cut down without losing the "quality" of the images?

Machine Specs --

P4 2.66, 512 MB ram

Render Settings:
A bunch of Area lights at quality lvl 4 (one for each floor and ceiling surface, and one for the window wall).

reflect: on, shadows: on, transparency: on, refraction: off, caustics: off -- thoughts?

[ddho1981]

2nd image:

[kevman3d]

Well, unless you need to raytrace volumetrics behind transparent objects, then deactivating 'Raytrace Transparency' might be a start...

Also, drop your Rays Per Evaluation to like 4 or 6 (rather then 16), and deactivate 'Raytrace Optimisations' as it can actually INCREASE render time.

Also, how have you set up your Antialiasing? 'Enhanced' modes will give slightly better quality then normal modes, however for still images you could almost get away with using 'Low', and use a small 'Adaptive Sampling' value like 0.1 or smaller.

However that said, you could deactivate 'Adaptive Sampling', drop the quality level of your area lights to like 2-3 and then activate under global illumination the option to smooth shading... This works pretty well for Radiosity, so I'll assume that it might work fine for grainy low-quality Area lights as well...

Finally, always do a quick test at a low res for any suggestion and see if it might work before commiting to a high-res final...

BTW, nice images! and good luck! :)

[js33]

Also your perspective is off on the second image. What is your zoom setting? Increase above the default of 3.2 to 4.5 or 5 and move the camera back. Your sink is blown out as well. Drop the diffuse and up the reflection. Are you going to composite something in the windows later?

Other than that very nice images.

Cheers,
JS

[hrgiger]

In the render panel, set your ray recursion limit to a lower number then the default 16. That's basically how many times your reflection is going to bounce so if there are two mirrors facing each other and you were between them, your image would only be cast 16 times instead of infinitely like in real life. Even a lower number like 2 or 3 will suffice.

[jutland]

You also might check number of lights and thier intensity. Large numbers of lights add to the render time.

[ddho1981]

Thanks for the tips guys, i'll play around with what you've said and post the results -- btw, I'm new to the group but have been watching it for a while now and am glad to be on board...(no pun intended)

D

[ddho1981]

These are reduced versions of renders taking into accoutn what you've all said ... i was able to reduce the render time by more than 50% so thanks! but i'm getting some funky splotchiness i think from area lights . . . there's still a model-fixing i need to do here and there . . . any further advice?

[ddho1981]

and the 2nd one . . .

[WizCraker]

Use shading noise reduction in the lights it is in the same place you turn on/off radiosity.

[ddho1981]

those images were rendered with shading noise reduction . . . i think my light quality is too low so i am increasing it for my next render.

[ddho1981]

Originally posted by jackcarter
there is a thing that i haven't yet tried but it said to put a 100% transparent box in front of the camera, i don't know what it does, but it is supposed to take 50% off the render time

Yeah, I've read about that -- though i don't think it allows for shading noise reduction.

anyone know?

[trentonia]

OK...here goes! You can go to http://www.respower.com where they have a feature you can use to split your single image into a number of pieces (that you define), and spread the render accross the entire farm. If you're NOT using radiosity, then your 9 hour render would run in about 5 minutes at a cost of about $20.00, while the 20 hour render would take about 12 minutes and run a little over $40.00. If, as you say, you've been able to cut your times in half, then the times and costs that I have listed will also be cut in half. If you ARE using radiosity, you could try the "transparent polygon in front of the camera trick" (do a search on this forum for the threads where it is discussed). If you do use the farm, be aware that you can split your scene and render in any format, but only the .FLX format has the ability to be restitched. All other image formats you will have to stitch yourself. So, render your file in .FLX, have it restitched, pull it into LW (where you can avail yourself of the many adjustments allowed for .FLX files), and save it out of LW in any format you choose. Cost a little money, but saves a huge amount of time.

[WizCraker]

Why pay someone else to render your images when you can do it yourself. It would get mighty expensive everytime you want to render out that new cool model if you keep going to respower. Or do you work there and like pluging respower to get more money for them?

[trentonia]

No, I don't work there. I'm just trying to help the guy with his problem. It's not that expensive especially if you pass off the costs to the customers. Get use to the idea, because render farms are going to be a necessity in the future. Even though computers are getting faster and faster, animators see that as a cue to make their scenes more and more complex. In other words, render times are increasing even on faster machines with more RAM. Just my theory. Sorry.

[kevman3d]

Originally posted by ddho1981
those images were rendered with shading noise reduction . . . i think my light quality is too low so i am increasing it for my next render.

I would also download and play with Eki's Plugpak.

http://www.kolumbus.fi/erkki.halkka/plugpak/

Otherwise with low-quality area lights, you can also try a test with more AA passes as well (and remember to *de-activate* Adaptive sampling). It *might* work but it might also increase render times - You'll have to test it and see...

Looking good!

Kev

[WizCraker]

Originally posted by trentonia
No, I don't work there. I'm just trying to help the guy with his problem. It's not that expensive especially if you pass off the costs to the customers. Get use to the idea, because render farms are going to be a necessity in the future. Even though computers are getting faster and faster, animators see that as a cue to make their scenes more and more complex. In other words, render times are increasing even on faster machines with more RAM. Just my theory. Sorry.

No need to be sorry, I was just saying for an image it can get pricey, but I guess if you past the cost to the customer it does not. Actually I was talking more about for not doing an animation or an image for a client but for self improvement where the cost can go up.

I do however believe eventually we will have close to real time rendering in the future. It only took them less than a decade to figure out how to do it on non complex geometry. But before this can happen I think the Industry needs to get together put who's better aside and come up with a rendering standard for interfacing with hardware. So if you use Maya, XSI, Lightwave, Max, etc... regardless of package xyz hardware render board will work. Modern Video game boards do this where in realtime you can manipulate geometry and textures, complex or not.

Now Back to the thread [if we continue this we need to open a new thread somewhere else]

[ddho1981]

personally, i'd like to FIRST get a handle on how to optimize my renders on my comp. . . . if need be, i will farm out--but knowing how to optimize AND farming out will produce much faster/cheaper costs later on when i'm tight on time. at the moment i'm not in a hurry so why pay for experimentation?

in anycase, here's the latest render . . . it took 4h 29m 5s:

comments, concerns, critiques?

I'm going from 512MB of PC133 ram to 1GB of PC2700 by the end of the week . . . hopefully that'll give me a boost.

[trentonia]

Hey, here is the thread that discusses the transparent polygon trick. This is supposed to speed up radiosity renders by a factor of 2. http://vbulletin.newtek.com/showthread.php?s=&threadid=9683&highlight=transparent+polygon+radiosity