Decreasing render times


This is the first year the team I’m in is entering an animation. The thing is, though we’re doing an animation which I believe is quite good considering our skill level, we’re looking at 4min 30s render times, which we’re going to need to cut down. Any major suggestions? We are using mental ray with final gather (on what amounts to low default setting), and no funky stuff like Hair and Fur etc. I tried using scanline, but it came out very dark and awful. Any suggestions?


P.S. A kind of related question, why doest the daylight system show up on mental ray?

It all depends on what you are trying to achieve, I use a layer pass render technique. Its what all the major animation studios uses too. You render out almost everything out in a separate layer and composite it in a third party program, Like Combustion (which is legal). This way you can achieve soft shadows, motion blur, distance blur, High quality Global Illumination on the backgrounds, and blurred reflections. Its by far the best way to render because then you can tweak your footage and get instance feedback without waiting 20 or so minutes for a result. The problem with this tatic is that you have to know what you doing when you composite, it has a little learning curve to it. (as with any other thing in 3ds max)

Here is a tutorial to give you more insight on the subject. Rendering in layers

I highly recommend you learn this way of rendering. It saves you tons of time in the rendering process.

Now what if you don’t have a third party program or dont have the time to learn the program and learn layer rendering? Then you really need to look at your max file. Do you have more lights in your scene than what there needs to be? Are there some objects that have ray trace materials on them that really dont need to show reflections. (example: a highly polished wood furniture in the background) You need to go through your file and see stuff that is slowing you down. I would recommend sticking to mental ray. Scanline tends to be really slow. And your test renders should not involve complex lighting calculations. They are only to render out the movement so you don’t need to see all the reflections and soft shadows.

Now after you’ve gone through your file and fixed everything there are one of two things that you need to do. You need to sit down and ask yourself, “the way this render time is going, do I have enough time before the deadline to complete rendering?” If the answer is yes, then you are good. If the answer points to No here are your two options.

1: Make sacrifices, first thing to go should be motion blur. Then distance blur, and then soft shadows. Get rid of objects that are causing render time to increase. Its very painful but sometimes its what has to be done.

2: Get more computers. Have a Render Weekend and gather as many computers that you can and have them all render a piece of the animation. You have to load 3ds max on each of them and set the paths to the textures and all. Or if you know networking computers you can run backburner, which comes with 3ds max.

I have considered network rendering because I have a book explaining how to do it. I think I’ll go through the file first though and eliminate stuff :frowning: . Thanks for your help.


I don’t fully understand the Mental Ray renderer but I noticed the Draft preset rendered much quicker than the High preset.

So I created a new preset called “Mental Ray Medium” and tweaked the settings to fall someplace between High & Draft. I experimented a bit to get the best compromise between speed & quality.

We ended up with render time about 1/3 of Mental Ray High with only a slight loss of quality.

Note that you do not need more than 1 license of 3ds max to use multiple computers as render nodes. You need 1 licensed copy as the master renderer to send assignments to the render nodes, but each render node can have an unlicensed copy of 3ds max installed which allows you to use as many computers as you can find as render nodes (on the same network…)

wow, it’s been a long time i’ve thought about animation…
as i recall, it’s best to use low values for the MR renderer for test rendering and such, but higher values (if you can wait) when it comes to the final render. Sometimes there’s no noticeable difference between high values and low values. I usuallly start with values below 100. If you are simply test rendering to see how your lights will appear in the scene, it’s best to go low because you just want to see where your light cast shadow and the coverage. what are your final gather values?
also consider, does your whole animation need to be rendered with final gather? can you compensate by not rendering with final gather in some scenes and still achieve relatively the same results? just some thoughts…

Well, I’ve reduced everything down, and I’m now at about a 3:30 render. Doesn’t look quite as good, but oh well. I turned all the settings up to what I wanted them and just for jokes and ended up staring at my screen for nigh on 15 minutes, so all for the best I guess…


You have a week, get on it this weekend and let your computer/computers render the entire week. I suggest you render out in a targa sequence file and combined them all in a third party program so you do not lose all your data if your computer crashes.

Oh and FYI: Your animations will never look as good as you want them to be. It’s just something you have to deal with. Every time I look at my animations I always think “My goodness!! That looks awful! I could have made it look better.” Even my personal project right now I think looks god awful. I’ve even talked to professionals and they say that they always cringe when they see their own work. The problem is that you see your work as what it should be and not what it is. A fresh pair of eyes, who knows animation, will be able to tell you if its good or not.

If you do look back at your previous work and do not cringe, that is a BAD thing. It means you are not growing as an artist.

Anyone want to enlighten me on how to network render?

I have a guy that’s bring in a CPU with 16 gig ram and some quad core $@#$@#$@#$@#. The computers we’re running are already pretty pwn… so if we could network up with the ultra pwn cpu it would be a huge help.

Here is the best way of network rendering. You be the network. Go around and assign each computer specific frames and have them render it. Then take the video clips and combined them in a third party program. If you dont know how to set up the network now, you probably want to play it safe and not take any risk because its getting down to the wire.

remember that one time you doubted when scheiße was down to the wire and you said for me not to try it… then i did… and it worked… and it was awesome…

dont doubt me bro…
dont doubt me…

I’m not doubting anyone, If you believe you can do it, go for it. If you have absolutely no idea how to get computers to speak to each other, you may want to stay the safe route. Either way, do what is in your ability, there are only 13 days left to have a complete animation. Its not a lot of time to be dabbing into unknown territory.

well i got our electrical and programing team to hook some scheiße up. so its cool now.

:eek: <–devon likes that smiley.

You know… being almost down to the wire with the animation, and STILL have a lot to do because of some local problems, I’m just sooooo glad that everything’s kept simple in this animation, that I can render the whole thing (if it gets completed) in a couple of hours or less.