Flythrough Questions

[Burnsyman9000]

Hello, after reading the flythrough tutorial in the help files I have a few questions. Any help or links to relevant threads is much appreciated.

The tutorial shows the LC and IR precalculations being done as separate backburner jobs. First the LC then the IR. I've always done the LC and IR precalculations at the same time as part of the same job and it seems to work fune. When submitted as the same job it first calculates the LC then the IR. Is it wrong to do it like this? Am I getting incorrect results by not submitting them to backbruner as separate jobs?

The tutorial mentions that the IR precalculations can't be network rendered over multiple machines and should be handled on a single machine. Is this still the case even with current Vray builds? I have been submitting the LC and IR precalcs to 3 network machines and it seems to work fine and also speed up the precalc. It looks to me like different machines are handling seperate Nth frames in the incremental IR calculation and adding them to the IR file, but maybe something incorrect is happening that I'm not aware of. The animations look fine in the end doing it this way but would they look better if I just did the precalc on a single machine?

The tutorial suggest that the LC precalc should be faster than the IR precalc but I have the opposite by a large margin. The LC portion of the precalc is taking over 8 hours for a 40 sec sequence using 15000 LC samples. the IR portion of the precalc is taking only around 40min, using 120 Hsphere subdivs and every 10th frame. Is there some reason why the time required would be the opposite for me of what the tutorial says I should expect?

thanks again for you're help!

[joconnell]

Definitely something wrong there - it could be a bit of bad geometry or overlapping faces. I know it's a pain in the ass but if you can try gradually hiding bits of your scene to try and identify what part is causing the slowdown you'll find the object causing the issue.

[peterguthrie]

15000 LC samples? Isn't that really high?

[Pictor2]

15000 is really high, especially if it's only 40 secs. 3000-4000 max should do it. You may get away with as low as 2500 depending on the size of the scene.

[Burnsyman9000]

is 15000 samples really high? I thought that was spread across all 1200 frames meaning roughly 12.5 lc samples per frame no? maybe I don't really get how that works in a flythrough - so much to learn! Any pointers on the first 2 questions? Thanks so much for the help!

[titane358]

is 15000 samples really high?
yes, and u will have some problems for rendering as the LC file will be very heavy.... bad idea.

[Burnsyman9000]

Seems to render fine and actually LC file comes out to around 60MB in the end so not too heavy. The animations look really good, but it's nice to know I can try reducing the LC samples without too much of a hit in GI quality. Will run some experiments with lower setting to see what I can get away with. Does anyone know if it's fine to precalc the IR and LC at the same time, and not separate jobs like in the tutorial?

[Pictor2]

If you mean calculate them simultaneously then no I don't think so as the irradiance map uses the LC to speed up its calculations but you can get the irradiance map to start calculating automatically after the LC has finished by checking the box 'switch to saved cache' in the light cache parameters. This way you can do the entire pre-calc in one go.

Summary - Primary Bounce Irradiance map, Multiframe incremental. Secondary bounce Light cache, Fly-through mode with 'switch to saved cache' enabled.

Note this is for static scenes with only the camera moving.

[joconnell]

is 15000 samples really high? I thought that was spread across all 1200 frames meaning roughly 12.5 lc samples per frame no? maybe I don't really get how that works in a flythrough - so much to learn!

I'd tend to think about the are you cover rather than the amount of frames you're making the lc over. For example it doesn't dump all the samples from frame 1 and then remake everything from scratch in frame 2, each sample is actually a 3d point in space where vray has measured the lighting so if you've got a slow moving camera, the exact same sample might be visible and valid for really long sections of the shot. If you had a really fast camera flythrough then your samples would be further apart alright and lead to some slightly patchy looking GI - if you've got a gentle enough camera move though you can probably use less samples and still have a nice smooth result.