Clarification Needed...

[sv]

Hi,

I've been using vray for some time now, but i have never actually be taught it. We saw the quality of vray work, and decided that we needed to use this render in order to get better at what we did. Keeping this in mind, can i just get these questions and queries answered:

Is this correct: As vray is an addaptive renderer, it will take more samples automatically when it needs to, in order to get the desired level of quality. Thus, by setting the minimum qmc samples to a higher level, such as 16, 20 etc., you could leave the thresholds for things like Irradiance map to be quite low, (as these would represent the minimum number of samples to be taken), but vray would automatically adjust itself where needed because the higher number of qmc samples would effectively override this.

Material Interpolation: When using lightCache, you can choose to use the LightCache data for all the glossy rays - not the default qmc method, thus getting faster renders. Under the vray material settings, there is an option to turn on interpolation for reflections - is this effectively like using the Irradiance map data to compute these rays - i.e. not using the qmc, brute force method?

Not sure if either of these are true - i'm just trying to understand what the hell that piece of software is doing, and hopefully get faster renders out of it.

Thanks to anyone for any help!

[flipside]

Interpolation for glossies is indeed like IR map for GI. It uses similar settings and tries to approximate the true glossies by rendering in passes etc... But I would not use interpolation as it usually doesn't give good results and you will be better off by simply lowering the subdivs value for the glossies.

Vray is adaptive and you control how much and where it should adapt. You do this globally in the QMC sampler rollout (these settings affect all qmc related calculations: ir map, dof, mb, vray shadows, glossies, ...). You can also adjust everything seperatly like for example the IR map controls. The ir map thresholds are used to detect areas where vray should take more samples for the ir map.


Don't know if that was an answer to your question, I didn't understand it very well.

kind regards,


wouter

[sv]

Don't know if that was an answer to your question, I didn't understand it very well.

It did in part, i know it didn't explain it very well.

Basically, interpolation for materials - great, this answers my question. I think interpolation should only be used when the glossy values are high, and you will not be seeing ny detail anyway - such as concrete with 0.7 glossy values. In this case interpolation is much better than lower subs, as you don't get very noisy results with quicker render times - you loose detail, but you wouldn't see it anyway. For instance, this is needed to get specular light highlights, but no detail is needed.

about using lower quality settings, but having higher qmc samples. what i meant is; to get good results but fast renders, could you set the Irradiance map to very low, such as -5 / -2, but set the qmc samples to high - say 16. This way, low samples would be taken where there is no detail, but the high qmc value would mean that higher values would have to be taken in areas that need detail. Basically, (not sure if it works this way or not) it would just eliminate tracing samples in areas where you don't need the details.

Not sure if i have explained this right, but thanks for taking a look & answering

[flipside]

I assume you're talking about the qmc multiplier in the qmc sampler rollout?

This multiplies qmc values, like the hsph in IR map settings. It will not add more samples, it will increase the quality of each taken sample.

What you need is the ir map thresholds. Do tests with show samples turned on to see how these thresholds affect sample placing. Especially the clr threshold is very important and sensitive (very sensitive around 0.3 value).

If you lower the min/max, you will loose detail which you will not gain back by increasing hsph or qmc multiplier. If small GI detail isn't an issue, lowering the min/max rate is a good idea. If you then use higher hsph subdivs you will make sure that the low detail GI will look clean. Don't know if that makes sense

Best thing to do is making a small testscene of an interior with some high and low detail areas. You can get similar looking renders with similar times by using completely different settings.

[flipside]

Now that I read your post again, I think you're confusing IR map with QMC GI. If you use IR for first bounce and QMC for secondary, the value of QMC GI has no effect on the IR map first bounce I think. Not sure about that.

But since QMC is set for secondary, it will not add detail to your first bounce since that it is taken care of by the IR map. If secondary bounces take up a large part of your total lighting, it will of course look as if increased QMC is affecting GI quality very much, but this is only the secondary bounce quality. But in these cases you would be much better of by using lightcache for secondary bounces.

[sv]

Flipside - cheers for this discussion.

Basically i'm aware of how each gi method works, and your info has certainly cleared up a few thoughts. What i was wonderng was whether high qmc values would affect the irradiacne map solution - but basically no.

Good settings i have messed around with are min/max rate at -5 and -2,(which is low) but upping the value of the colour, normal and distance thresholds to 0.3, 0.1 and 0.3. I find this is quick, and you still get good results.

What i was wondering was not to do with secondary bouces using qmc, but whether a higher Min Samples value under QMC Sampler would mean that you could set the Irradiacne map to be low settings, but this higher QMC value would cause extra details to be taken. From your posts, no it seems is the answer. I wasn't sure how much of an overlap the QMC sampler had on other GI elements.

Basically i want to find a way to reduce the amount of minimum samples to be taken for the lower quality samples in the Irradiance map, but keeping the higher samples at a decent level. I wasn't sure if the QMC sampler would affect this.

Thanks for your help.

[flipside]

Note that I'm not sure on the effect of QMC for second bounce on IR map solution.

The min qmc samples in qmc sampler rollout is something I never use, because I can't figure out exactly what it does.

The min/max rate is resolution dependant, so -5/-2 can be very high, for example on a 6000*4000 pixel image. I always do test renders at 640 to 800 px wide images, and when I found satisfying settings (the GI detail/quality is good) and want to render larger, the only thing I change is min max rate (lower each by 1 for each doubling of resolution).

what you say here:

Basically i want to find a way to reduce the amount of minimum samples to be taken for the lower quality samples in the Irradiance map, but keeping the higher samples at a decent level. I wasn't sure if the QMC sampler would affect this.

is exactly what the ir map thresholds do. What you could try is a higher clr value and a higher dist value. Because the clr is so sensitive, when you go from 0.3 to 0.4 for example, vray will take A LOT less samples, decreasing rendertime drastically. To compensate for the loss in detailed areas, increase the dist value (higher dist means more samples, while the other two thresholds work the other way around).

Here's what I mean:

clr=0.3
nrm=0.3
dist=0.3



clr=0.4
nrm=0.3
dist=0.5



Samples off with first settings



Samples off with second settings




As you can see, the second one is much more efficient. This reflects in rendertimes: 14s for first and 9s for second.

Of course it depends on the scene, but for interiors this can be usefull. Clr th is especially needed when you want to see GI shadows casted by small objects.

The images are rendered without secondary bounce engine.

[flipside]

Ha everytime I try to answer, and read the question again after I post, I understand better what you mean

I don't think there is a way to use lower quality for samples in low detail areas. It would be nice to be able to use lower hsph values for samples in non-dense areas though.

[sv]

It would be nice to be able to use lower hsph values for samples in non-dense areas though.

I know, i find a lot of the time loads of CPU power and time is spent on lower quality samples. It's hard getting the balance right sometme.

But thanks a lot - this is really useful.

Clr th is especially needed when you want to see GI shadows casted by small objects.

This is what i want to achieve in final renders, but have previously just stuck it on the default High setting, and waited for a while. We had a really nice image that was looking out of a window across a cityscape this week to render. Took 4 and a half days - one image, 3500pixels wide - all glass, but only on one machine because we can't get DR working just yet. Was absolutely pi**ing myself we woudl have a power cut!

Cheers.