Photometric Lights = Long Render Times?

[illegalalieninbeijing]

Hello Mental Ray Faithfuls,

 

I need expert advice on dealing with lengthy render times I am having on a couple of interior scenes I am working on. Render times can reach over 20 hours on a single 2400 X 1560 ppi image. It is easy to point that the culprits may be the 50 photometric lights placed inside the model that I cannot do without. Without the photometric lights and using exposure and gamma control instead to illuminate the scene, the same image would render just over an hour. But of course, the result would be less desirable. But I am not eliminating the possibility that I could have set my GI, FG and other stuff wrongly thus making things complicated. Or perhaps my render times are normal for these scenes?

 

Anybody to the rescue and be my hero?

 

I have attached 2 of the WIP scenes and most of the settings. Hopefully, these are enough to give you an insight what’s going on. Let me know if you need more information about the model.

 

About the model:

- Modeled in Sketchup Pro 6 and exported to 3DS format.

- Polys: 728,871

- Verts: 521,433

- 20 free rectangle PL, shadow ON, uniform diffuse, 25%

- 22 free rectangle PL, shadow OFF, uniform diffuse ( pointing upwards), 22%

- 6 free light, shadow OFF, spotlight, 20%

- 2 free light, shadow OFF, uniform sphere, 40%

- 1 daylight system, shadows ON

 

My Machine:

- Autodesk 3ds Max Design 2011 64 bit

- Windows 7 Professional, Service pack 1,

- Intel Core i7 CPU 960@3.20 GHZ

- 12.0 GB Ram

- 64 Bit Operating System

 

Thank you very much for all your help. Any suggestions and ideas will be greatly appreciated.

 

[ATTACH=CONFIG]42477[/ATTACH][ATTACH=CONFIG]42478[/ATTACH]

 

 

[ATTACH=CONFIG]42476[/ATTACH][ATTACH=CONFIG]42475[/ATTACH][ATTACH=CONFIG]42474[/ATTACH]

 

Due to the limit in attachments, I have attached the next 5 of 15 on the reply that follows.

Edited April 28, 2011 by illegalalieninbeijing

[illegalalieninbeijing]

[ATTACH=CONFIG]42488[/ATTACH][ATTACH=CONFIG]42487[/ATTACH][ATTACH=CONFIG]42486[/ATTACH][ATTACH=CONFIG]42485[/ATTACH][ATTACH=CONFIG]42484[/ATTACH]

[illegalalieninbeijing]

[ATTACH=CONFIG]42493[/ATTACH][ATTACH=CONFIG]42492[/ATTACH][ATTACH=CONFIG]42491[/ATTACH][ATTACH=CONFIG]42490[/ATTACH][ATTACH=CONFIG]42489[/ATTACH]

[pixelperfectg]

Couple of tips:

1. Use the Far Attenuation feature of the photometric lights to limit the sampling range. This feature, when configured properly, can substantially reduce your render times when using lot's of photometric lights.

2. I would never use the same image sampling values for both min/max settings. You're not allowing mental ray to use adaptive sampling to speed up the render process. Right now it's using the same amount of sampling over the entire image whether it needs it or not which isn't efficient with mental ray.

[illegalalieninbeijing]

Couple of tips:

1. Use the Far Attenuation feature of the photometric lights to limit the sampling range. This feature, when configured properly, can substantially reduce your render times when using lot's of photometric lights.

2. I would never use the same image sampling values for both min/max settings. You're not allowing mental ray to use adaptive sampling to speed up the render process. Right now it's using the same amount of sampling over the entire image whether it needs it or not which isn't efficient with mental ray.

 

Hi Jeff,

 

When I was doing partial test renders, my original intent was to use the Far Attenuation feature of the Photometric Lights. But it seems that I couldn't get pass the Final Gather computation. But I'll give it a try again tonight.

 

I agree with you on the image sampling. I ended up using these figures because it seems to render faster. My intent was to have it at min. 1 and max. 16 or 64.

 

Thanks a lot for the input. Please feel free to comment on it some more. I will post the results soon.

[BrianKitts]

I used to overuse IES and then one day I realized I was using them in places where you don't even see the benefit. They're best used as an accent light, if you're lighting your whole scene with it.....say 60 IES lights then you're just throwing useless havoc into your render times. Unless you're doing a photometric light study, there's just no need.

 

Lighting is artistic, set it for your desired result and do it in the best way possible. This means only use IES when it's again a wall and you're really going to see the attenuation and caustic rings imbedded in the file. Otherwise stick to standard and vray lights (or whatever your engine you use) and the time you save in rendering will make your life a heck of a lot easier.

 

I look at your images, and I don't see a single case for IES to be used at all......

Edited April 28, 2011 by BrianKitts

[Justin Hunt]

When setting the far attenuation you can also lower the shadow samples, which are real render time killers.

 

jhv

[Crazy Homeless Guy]

Also..

 

It is important to remember that all IES lights are not equal. Some generate from a single point, others from square/rectangle/round area lights. The area lights can slow down your rendering quite a bit. So, you might want to play around with overriding the area lights, and setting smaller ones, or in some cases, turning them off altogether.

 

Also, I believe that the number of scallops and the shape in which light is thrown can have a fairly dramatic effect on render times.

 

I would hihgly recommend setting up a scene with a blank room and 3 IES lights. Then switch out diffferent IES files to see how the render times differ, and how the light being thrown from the sources differ. This will help you quite a bit in making decisions about how to use them in your scene, if you use them at all.

 

I fall into the camp of believing that we do not need to create everything in a perfectly accurate way. We should use techniques which produce the results we desire. By this I mean take liberties. You don't have to use the exact IES that the manufacturer provides just because it is the one that actually matches the specified light in the scene. Unless you are doing actual light distribution calculations, no one will notice.

 

Also, doing things like not making lights effect glossy reflections/refractions can reduce render times. Depending on the proximity of the light to the camera, I may even turn off effect specular reflections.

[Camby1298]

I fall into the camp of believing that we do not need to create everything in a perfectly accurate way. We should use techniques which produce the results we desire. By this I mean take liberties. You don't have to use the exact IES that the manufacturer provides just because it is the one that actually matches the specified light in the scene. Unless you are doing actual light distribution calculations, no one will notice.

Funny Enough, Most of the time when you do exactly how the designer/architect desires, they don’t like it, or rather would like to change it (they're designers, that’s exactly what they are supposed to do). I find that no matter how much they (client/architect/designer etc.) insist the design is final and there are no changes needed to be made; as soon as they start to see their design unfold in front of them through the use of progressive updates; they always change the design. Always. But that’s the nature of the beast and it only makes sense that this happens. It's extremely difficult to visualize every detail of a design 100% through one's imagination; and is one of the many reasons why we (architectural visualists) are here.

Typically if I’m setting up a scene/render as specified from the designer, I always try to set up options and tools to accommodate design changes if they should come up, so there is little stress to meet the deadline.

As Travis said, don’t get stuck on getting the exact numbers from a manufacturer's product, instead know what you and the designer want to see in the rendering and gear your workflow towards that.

[illegalalieninbeijing]

Hi All,

 

Thank you for all your prompt replies. Following most of your ideas and suggestions over the weekend, the render time was greatly reduced to about 3 hours for each scene. It was a throw off between here and there but It came out as intended in the end. I even manage to increase the image size, sampling quality, FG and noise filtering which boosted the sharpness of the image. There was no drastic change in render time which was still about 3 hours.

 

Again, thank you very much. I really appreciate it.

[Arnold Sher]

...use shadow map... Will speed up render times dramatically

[Justin Hunt]

Try turning shadow casting Off for the water.

 

jhv