V-Ray Render Optimization - an in-depth Guide (call for Before/Afters)

[AlexP]

Whats the best way to approach scenes with lots of detailed diffuse textures? carpets, woods etc.?
I guess you want the DMC using full power on those textures, but this tends to get unbearably slow.

[Dariusz Makowski (Dadal)]

Whats the best way to approach scenes with lots of detailed diffuse textures? carpets, woods etc.?
I guess you want the DMC using full power on those textures, but this tends to get unbearably slow.

U just have to spend the sample in order to get smooth effect. There is no cheat or short cut. Any complex textures should be orange/red. The best setting will mean AA will spend almost 0 resources on light/glossy rays.

[Rhys]

Why would that make a difference? They're not mutually exclusive.

Irmaps are significantly faster than brute force, and optimizing reflection/AA subdivisions doesn't change the IR map at all. My render times are 3-5x higher with brute force, and noisier (although i've not used 3.0 yet, while it's faster I still dont think it's fast enough).
It'll be a long time before I'm willing to take that speed hit.

I find at the moment I still do the optimising for AA and reflections, then just switch on the irradiance map. the results may not as be as nice but the speed hit of brute force is not worth it for all jobs!

[AlexP]

U just have to spend the sample in order to get smooth effect. There is no cheat or short cut. Any complex textures should be orange/red. The best setting will mean AA will spend almost 0 resources on light/glossy rays.

I see. In which case the only short cut is to buy more render power !

[RockinAkin]

Yep, there certainly comes a point when you've optimized whatever you can, and you simply need to give the renderer the time it needs to do the job properly.

Regarding GI:
For the tutorial I tried to keep topics as simple and straight-forward as possible, hence sticking with BF + LC. For other GI methods (cached), a whole other tutorial could be written.
So the workflow will still work with whatever GI method you're using, but for the GI part of the procedure you'll have to adapt to the settings and requirements of whatever GI method you're using.

[Neilg]

There's no long writeup needed. Drop reflection depth to 1, turn off glossy effects, hide all the glass objects, use medium 140/70 and save it out. that's going to work in 90% of situations for stills.

[cheerioboy]

As I work with this method of optimizing, I'm still looking for how to back down once fully optimized. I'm working with detailed interiors that will take quite a long time with the lighting & reflections balanced.

I ought to start digging into the progressive rendering in vray 3.0 - setting a time limit and then pushing the optimizing will be the way instead of optimizing and then letting the render go way over the desired time

[RockinAkin]

Once you've found a good balance for the Image Sampler (AA) and DMC Sampler... you can increase the Image sampler's color threshold value to speed up the render (of course this will re-introduce noise).
So it's just your call how much noise is acceptable.

[Donald2B]

I usually just adjust the DMC sampler noise amount to something high, like 0.2. It will produce a lot of noise, but renders are usually really quick and good enough to get the general idea of what's going on. It's also a nice way to let clients know that it is NOT a final render and more for a quick study or progress report.

[cheerioboy]

Thanks guys

So when 'use DMC sampler thresh' is off and you're using the Clr thresh, you simply adjust the Clr thresh. But if you're using the DMC sampler thresh for both then you dial that. No need to dial both when using the clr thresh for the image sampler. Thanks for the tips, almost there

[Fredrik Averpil]

One word: awesome tutorial.

(okay, that's two words)

[ben_hamburg]

So I´m back with some more troubles...so far your tutorial greatly helped me in improving rendertimes AND reducing noise...in one scene we had render times of 30-40 min per frame for 720p, now we´re rendering the same scene in 4240*1080 in the SAME time.
At least during testing.
For testing purposes I used Irradiance map and Light cache in single frame mode.
For animation I was rendering an animation prepass as usual (since we have camera movement and object movement), but once I was switching to animation (rendering) mode, render times exploded 10 fold.
I was using the "Low" preset for Irmap and default settings for light cache with use light cache for glossy ticked.
Since the scene is heavy on detail (lots of foliage, ocean with waves and reflections), I upped the DMC sampler to 1/32, with various materials up to 64 samples and one HDRI Domelight for the lighting with 128 samples.
Any Idea why my render times explode like this?

[joconnell]

I was using the "Low" preset for Irmap and default settings for light cache with use light cache for glossy ticked.
Since the scene is heavy on detail (lots of foliage, ocean with waves and reflections), I upped the DMC sampler to 1/32, with various materials up to 64 samples and one HDRI Domelight for the lighting with 128 samples.
Any Idea why my render times explode like this?

Making a jump up to aa of 1/32 is pretty damn high, I'd try scaring the crap out of yourself and definitely up your domelights and material even more than you have already - try going in multiples of 2 each time. Small changes need not apply

[ben_hamburg]

Well, I tried lowering AA to 1/4, which it was before, and render times are down to 1 hour per frame. Still bad...
I must confess, even though I read the tutorial over and over, I´m still not sure I quite got the reasons why those things are as they are.
Of course I can try upping subdivs on my light and materials (the latter beeing a pain in the a*** since mostly everything in the scene is either xrefed or proxies), but I´d just like to know why render times are so much higher during animation, compared to rendering stills...?

[ben_hamburg]

And like I said...I optimized the scene test rendering still frames, and there upping the AA samples to 1/32 was the sweet spot, lower or higher values resulted in slower render times.