Animated IMAP takes up to 10x longer than non-animated one

[MoonDoggie]

Friends,
VRAY1.5 (non-SP1)
I am rendering a scene and I did it two different ways with much different results:

- The first way I rendered every 20 frames of the imap and saved it out
- The second way I did an animated imap (prepass), and slightly animated the sun and saved it out

Here are the results rendering the same frame one from the every 20 frames imap, and two from the animated imap using different interp frame values of 3 and 2

Every 20th Frame - Render time: 6m 30.9s
Interp frames 2 - Render time: 54m 21.6s
Interp frames 3 - Render time: 1h 18m 43.8s

Same settings except the saved imap...Am I missing something?

Many many thanks,
-Colin

[vlado]

What are you using for the Interpolation samples of the irradiance map? Is it the default (20) or you have increased it? Also, what is your samples look-up type set to?

Best regards,
Vlado

[MoonDoggie]

Interp. samples: 40

Sample lookup: Density-based (best)

[vlado]

You can try to reduce the interpolation samples; it will speed up the rendering quite a bit.

This is because, in order to keep the blurriness comparable when loading several irradiance maps, V-Ray will multiply the interpolation samples by the number of loaded irradiance maps. When interpolation frames is 3, this means 7 irradiance maps are loaded; so for each shading point, V-Ray looks up and interpolates from 7 x 40 = 280 irradiance samples... which is quite slow.

It might help to use the "Nearest" look-up mode too.

Best regards,
Vlado

[MoonDoggie]

Excellent! I'll do this. I literally just transferred my settings straight from what I knew looked good at 6m per frame, to the animated imap.

I don't think I need to re-render my imap do I? the imap_frames rendered in a decent amount of time (~9hours). What other spinners are tied to the new imap animation settings that I might want to know about?

[Aldaryn]

This is because, in order to keep the blurriness comparable when loading several irradiance maps, V-Ray will multiply the interpolation samples by the number of loaded irradiance maps.

This is a very, very handy info, Vlado, thank you!

I was just about to open a thread for this, since the render time increment gave us some headache here lately. I just couldn't figure out, why a render with animated irradiance maps takes that more time to render.

By the way, what do you mean by keeping the blurriness comparable?

best regards,

A.

[vlado]

You know how when increasing the interpolation samples, the GI becomes more blurred, and decreasing them makes the details in the irradiance map sharper - this is what I mean.

When several irradiance maps are loaded, the sample density is increased; if V-Ray took the same amount of samples, they would cover a lot smaller area, leading to sharper details than a single-frame map. To compensate for this, V-Ray increases the sample count automatically.

Best regards,
Vlado

[vlado]

I don't think I need to re-render my imap do I?

No need to do that; interpolation happens during the final rendering, so you can use the same irradiance maps.

What other spinners are tied to the new imap animation settings that I might want to know about?

I think this is the only caveat there.

Best regards,
Vlado

[Aldaryn]

You know how when increasing the interpolation samples, the GI becomes more blurred, and decreasing them makes the details in the irradiance map sharper - this is what I mean.

When several irradiance maps are loaded, the sample density is increased; if V-Ray took the same amount of samples, they would cover a lot smaller area, leading to sharper details than a single-frame map. To compensate for this, V-Ray increases the sample count automatically.

I see, thank you!
( I thought that the sampling is done for each frame separately, and then the separate frame values are averaged at the end, but I think basically it's almost the same thing, and this method would also mean a multiplied interpolation sample count. So anyway, thanks for the info. )

best regards,

A.

[MoonDoggie]

Perfect, many many thanks as always.