Vray settings for 1080 frames

[Scott Schroeder]

I've got to render out a good amount of animation at 1920x1080. Normally I render much smaller than that, so I'm trying to optimize my settings as much as I possibly can. Right now, my frame times are around 30-45 minutes with saved light maps. I'm using Vray 2.0 and Max 2011 with an i7 2.93 and 8 gigs of ram. Upgrading the machine is clearly not an option. If it matters, I am using linear workflow.

 

Currently, I'm using adaptive DMC with a min/max of 1/4. My clr thresh is .02. I'm using VrayLansczos as the image filter. In the settings tab, the DMC sampler is set to a .85 adaptive and a noise thresh of .01.

 

If I have to use a glossy reflection, the subdivs in the material are no higher than 16.

 

I'm getting a great result with the above settings, but is there any more optimizations without introducing too much more noise? During testing on a smaller frame size, a DMC of 1/8 was slightly faster than 1/4. Would that translate to the higher frame size? I know you can decrease settings as you go higher in frame size, but I'm not exactly sure which ones to toy with.

 

I tried to adapt as much as possible from the Vray Speed vs. Quality video from the Gnomon Workshop, but as I've said, I really haven't rendered many animations at 1080 before so I don't know a good benchmark time.

[Tommy L]

Thats pretty much just the way it is. 1080 is actually more than twice as big as 720, so your render times are not too bad, depending on the subject matter being rendered.

When discussing a project with a client make sure to add a premium for a commercial render service if 1080 is needed in a hurry. Otherwise just tell them it will take a while and put it on low priority on the farm.

[thinice]

Hi Scott,

 

Curious to know what kind of scene you are rendering. From my experience, 30-40 minutes for 1080 frame is not bad at all.

[Scott Schroeder]

Thanks for your replies. What I'm working on is for a industry trade show. It's actually the booth at the trade show that is getting all viz'd up. But, due to the timing right now (as well as contracts) I can't show anything. But, your typical arch viz interior would be a good place to start. Except this has much more diffuse lighting and no sun since it's a trade show booth inside a convention center.

 

I used DMC of 1/8 and got it down to 26:38 a frame, but the noise is just hovering above an acceptable level. Though, when post effects are added the noise is eliminated and I am going to use some render elements.

 

Speaking of 720, has anyone actually used the plug-in from Red Giant called Instant HD? I've read some good things, but I'd like to know how well it really works. If that works, then I can render to say 720, then use that to go to 1080.

 

Unfortunately, the farm is my one computer and the 1080 requirement was just now dropped on me. As we all have probably dealt with when dealing with clients. I think I may have a work around and any animated shot is actually going to be done in a blueprint style, which means no GI/Reflections/Refractions and the things that really tank render times. I've got those frame times down to 2-4 minutes for 1080.

[thinice]

I do not use Vray but yeah, it is the GI/reflection combination that really kills it. I hardly ever use refractions when doing animations.

 

By the way, there are a few very good remote rendering services available if the client is willing to pay. They are not that difficult to use once you get the hang of it and actually not that expensive. Nowadays I use my in house farm (Q6600s) only for testing.

[Crazy Homeless Guy]

What is the size of screen and viewing distance? There is a chart that gives size of tv, and distance viewed. Following that will tell you whether anyone would actually see the detail of 1080, so you might want to check that.

 

Truth is, I would probably go 720, and stick with it. It will allow more room for user error that is bound to crop up.

 

But I reality, how many actual frames are you going to render? With the time you are getting at 1080, the best you will be able to do is 48 frames a day, and that is providing you don't use your computer for anything else.

 

But the bottom line is ...you need to go through a render farm service if you are going to deliver a professional product. One machine won't cut it.

Edited June 28, 2011 by Crazy Homeless Guy

[Scott Schroeder]

Since it's for a trade show, it's being shown on a variety of displays, from computers, 42-62" TV's all at 120hz and a 9'x21' Stewart Filmscreen display. I'm actually going to get a look at all of the tech next week, so I'll be able to see all of them in action. But I'm pretty sure on a 9 foot by 21 foot screen you'll be able to see some detail.

 

I've pretty much optimized things down to 15-20 minutes a frame for 1080, 6-8 minutes for 720. If I decided to go to an online service, that optimization needs to happen anyways.

 

For now, I've re-worked the shots so I can render this out in a solid weekend or two, worst case where I have to rely on one machine. I've simplified the moves, many of which are now done in post. When the camera does have to move, it's the blueprint look which takes all of 1-2 minutes a frame to render. Knowing that this could happen, I've allocated all of July/August for rendering and post work. So I've planned ahead.

 

Online render farms are great, but they aren't a magical solution. If you put crap into them, you don't get amazing images back. So to make a statement that the only way to get a professional product is to use a render service is a somewhat misguided statement. Since you pay per core hour, you really need to make sure you've optimized your shots and settings or else you'll pay more for needless rendering. So if I don't get these optimizations down, either I spend a month(s) rendering on one machine or bankrupt myself paying for needless core hours with an online service.

[Crazy Homeless Guy]

If a single computer works for you, then it works.

 

I have been lucky enough to have a render farm of some sort for the last 11-12 years, so this means I have never actually used an online farm, so I perhaps I spoke to easily about that. But in reality, I am surprised you are able to do your daily job with 1 machine? I work on 1 machine, but I am typically remoted into at least 1 other machine for testing, and processing items that will tie up my production machine for a long period of time.

 

And yes, if you put crap into anything you are going to get crap back regardless if it is a local renderfarm, or a remote renderfarm. Proper asset management becomes a must if you are rendering on more than one machine.

 

So perhaps, the problem you are alluding to of putting crap in and getting crap back is somewhat related to your work methods since you are managing from one machine? Just a thought. If the remote farms aren't predictable, then I can't imagine they would stay in business?

 

It sounds like you work for yourself. Have you considered building a 5 blade rack server or something similar? I can't help but believe it would pay for itself fairly quickly by allowing you to increase your production.

[Tommy L]

It sounds like you work for yourself. Have you considered building a 5 blade rack server or something similar? I can't help but believe it would pay for itself fairly quickly by allowing you to increase your production.

 

quoted for agreement. You're tying your arm behind your back using only one machine. Even for stills.

[thinice]

Just to give a perspective,

 

1 month on single i7 equals to less than 4 hours (on the service that I use) and cost about USD 1500.(standard rate)

 

I am not a render service evangelist and whenever I use them, I always make sure the client pays for it. Just that nowadays, I can't imagine servicing my clients without that option.

[Scott Schroeder]

Maybe my post was a bit more snarky that it needed to be. I 100% agree that using an online render service is a great tool and option. However, there still needs to be optimizations that must happen, and you must render responsibly or else you pay so much more than you need. If you roll that cost into your billing, you'll be out of work since no one will pay for it.

 

I have no issues working on one machine because I've got my settings for Vray so streamlined for each testing phase that I never experience any delay. I also use Vray R/T to quickly set lighting, rather than constantly test render. Sure, if I could DR I would. But for now, I've got to make the best with what I've got and that's one machine. If I'm just checking quick looks, do I really need to render full size? Do I really need full AA? Do I even need reflections/refractions or do I need to use all 5 or so reflective bounces or can I get the same look for testing with just one?

 

While I worked in the industry, I got lazy with optimizing. Just throw more computing power at it. Now that I'm on a very restricted budget and only one good machine, I've had to take another look at things.

 

Case in point. For this project, I have kept the camera moves really simple. Dynamic, but simple. I'm sure we're all tired of hearing comparisons to him, but the cameras are like the ones from Alex Roman's videos for Silestone and Third and Seventh. So, these cameras are rendered out at 2x speed. For example, 180 frames would actually be 90 frames of max animation. Then, using time and frame blending techniques and After Effects, the actual camera speed is set with zero queues that it was slowed down. Total render time for the 90 frames was a mere 3 hours on one machine. If I use an online service, I just halved your cost and therefore can show a lower billing rate and win more contracts. As painful as it is to admit that cost has played the biggest factor, way above quality unfortunately, in all of the freelance work that I have bid on so far.

 

If you have a 300 frame slow pan, push, pull, etc, and you have to figure out how to render it, you need to really go back to the roots of being an artist. Do you really need all 300 frames? What other tools do you have at your disposal? Think far outside of the box. Kick yourself out of the comfortable rut.

 

All too often the solution in arch viz and CG in general is to just throw more computers at it. Screw optimizing, screw other techniques, let's just jam this thing through the queue. I ran the render farm where I used to work before the recession, so I became very aware of poor rendering techniques as well as how to balance 4-6 complete rendering jobs that all happened to come due on the same date. After that, I taught students lighting and rendering using max. I would go on epic, class long rants about how they must optimize if they want to survive. It pained me to see students test rendering on great machines, but waiting 10-15 minutes. I've always had the 1-2 rule. If you can't get your test render done in 1-2 minutes, you are wasting time and you need to rethink your settings.

 

Once I get this job done and do some number crunching, I'm looking into investing into a personal render farm. I've been looking at the Artie system from Renderstream. Because if you pay all of the time for online rendering, you are wasting money when you could invest in a small turnkey rendering solution. The online service would then be for emergencies or the "oh by the way we need this tomorrow" jobs. It all depends on how consistent the work is that's coming through the door. If it's hit and miss like it has been, then the online service is better. If it becomes much more stable, then the personal farm is more cost effective.

 

I just want to end by saying I'm not attacking what anyone does or anyone's techniques. I only know what I've seen and experienced. For me personally, I've seen using video post as sort of this technique that people are afraid of. I know too many people that have the mentality that it must render 100%, and no or the only post that is done is to put the frames together into an animation.

 

I'm working on getting permission to show some of the proof of concepts that I've done so you can get a feel for it.

Edited June 30, 2011 by VelvetElvis

[Tommy L]

Good point well made Scott.

I'd add a couple of things though. Im a lone freelancer and I too started out with one machine having had the luxury of a farm at work. But render power is kind of like a cell phone or a car. Once you've had one, its hard to imagine note having one. I'd also say that some jobs just need one. I mean they just need it. If Im working on high res work, animations, multiple jobs, multiple render passes, doing post etc, I need two or three workstations PLUS a renderfarm. Feels like Im pushing up-hill otherwise.

[notamondayfan]

This is looking from a different angle, but have you tired Neet Image - De-noise filter for after effects? It's pretty impressive, as it will allow you to render with some noise and then run the animation through AE and eliminate a lot of noise, without blurring the animation.

It works really well for CGI animations as the noise is pretty constant over the whole animation, so once you have built up a good profile, it works really well. It's definitely worth buying, as you will use it again in the future, and it also works well for stills too.

 

Other than that all I can suggest is to make sure all models, textures, maps are all streamlined as much as possible. Try to avoid glossy reflections. Also keep the number of lights as low as possible.

 

Regarding render settings, I would start with Adaptive DMC 1-100, clr threshold 0.01, area AA filter, adaptive amount 0.9, noise threshold 0.1

[Crazy Homeless Guy]

A farm still needs to be used responsibly. Throwing poorly managed and set up projects at it because it is there isn't responsible. It also won't necessarily make a project better given the tight deadlines that the industry demands, and for that matter, the deadlines that all of architecture demands.

 

I try to keep our files lean and efficient so that when last minute changes need to be incorporated, they can do so with out wreaking havoc on time lines, and causing unnecessary stress on everyone involved.

 

But yes, having a farm and extra machines does allow you to do things you wouldn't be able to without a farm. Sometimes this is a very advantages thing, and other times it is not.

 

Also, a farm can be used for rendering out your AE comps. Things like frame blending can take a very long time when using AE on one machine. If you set up an AE farm, you can distribute that job over multiple computers, and then compile the individual frames into a finished movie.