Understanding DMC Sampler

[adanmq]

Yes Vlado the samples hits are amazing, help me a lot to understand the engine. Thanks for it.

Kosso_Olli. Don´t forget about "min samples" inside "Global DMC" this value overrides the minimum samples, if you put 2 you can get 128/2048. But try it before you lower it to much. And play a little bit whit the margins, they are too low for my xD.

The only problem i´m getting using this method is that sometimes its hard to force Vray to get to the maximum and the less noise areas could be over-sampled. But in general i´m getting cleaner results in less time. I only try it in heavy GI interiors, will give it a try in exteriors.

[jasonhuang1115]

You get clean noise using 256, now, in the "brute force" tab, "Per Pixel" secondary samples, in this example 4096/4096, take note of this number.

Hi Vlado, is it possible to add this sample hit number per pixel feature for Maya so we can try this optimization technique?

[vlado]

Hi Vlado, is it possible to add this sample hit number per pixel feature for Maya so we can try this optimization technique?

It's not as easy in Maya, but we'll see what can be done.

Best regards,
Vlado

[jasonhuang1115]

It's not as easy in Maya, but we'll see what can be done.

Best regards,
Vlado

Thanks Vlado. Is it an UI design or coding issue or internally VRay in Max and Maya using different way for secondary ray sampling?

[olly]

Not sure if I am being a div here, but when I use Fixed, reset all subdivs to 8 (using the VMC script ), set MSR to 256 and Adaptive Amount to 0.0, my "Per Pixel" samples are listed as 25600/25600. Surely that is not correct? We are using 3.10.01 and max design 2014. Take a look:



Vlado, you may have already spotted this but, with larger numbers, the "px" samples drop down a line and become hidden in the VRayMtl UI.

[adanmq]

Hi.
Olly, can´t see the image it´s yo small.
In my example i was using AA=4 (16AA samples) and MSR=256 (256x16=4096) if you use AA=8 (64AA samples) and MSR=256 you should get 16.384 (256x64=16.384) using "Divide Sahding rate". Sometimes you need to move the Subdiv slider in order to refresh the subdiv Hints, sometimes they don´t show the real value.

[adanmq]

PD: jasonhuang1115 you can try it even without the Subdiv Hints, just use a normal calculator instead the Hints to get the "per pixel" samples information. If you use Fixed AA the calculations are very easy.

AA Subdivs² x MSR = Per pixel samples

[olly]

Sorry, I have edited the post to fix the tiny image!

That makes sense now. As you will see I am using fixed 10 so that explains the 25600. I think the problems come from using RPManager and the hints don't always get updated when you change a spinner - my bad.

Vlado, you should be able to see the UI formatting issue now though.

[AshtonWoolley]

WOW.. What a thread to rapid read in an hour.

I was aware of the old method of optimizing a scene - Adjusting each secondary ray source Subdiv until you get no noise in that render pass.

Are we now talking about a new method that is simply leaving everything at default and changing the Min Shading Rate and then adjusting the Adaptive Amount? Is this faster (In terms on render time) than the old method "Grant Warwick" and a few others initially discussed of checking each render pass for noise?

Very confusing which method to pursue.

[joconnell]

So!

It's kind of a toss up between the amount of materials in your scene and the time you have to optimise it. If you've got the time, doing stuff per material will get the best results but it's man hours. If you've less time or loads of different materials the new theory is to find out the cleanest and noisiest areas of your scene, get 2 sets of settings that work for the cleanest area (leaving the noisiest area bad) and then the noisiest area (leaving the clean area totally over sampled) and then using the adaptive amount and thresholds so that the minimum amount of sampling done matches the sampling for the cleanest area. You get the two extremes and let everything in the middle sort themselves out.

I was a big exponent of the sample rate method even though it's only showing one piece of the puzzle (like ideally we'd have a reflection sample rate and a lighting sample rate and so on) and another big thing is that I'm passing on renders to compositors that might do quite large changes in exposure on each of the elements, thus they need really clean passes so they can push them around a lot. If you've got a really long render or a load of scenes that use the same materials though, what'd be more efficient again would be to render clean passes, let comp do all their balancing of the elements, then try and replicate what they've done to the materials back in the original 3d scene. It'll make things more realistic as compers will break the energy conservation of materials easily, and it'll also mean that vray is only doing the work it needs to so the beauty is perfectly clean, and not for the beauty plus a load of extra quality to allow pushing around in comp.

[jasonhuang1115]

John, any suggestion on using the "new theory" in Maya where the info of min/max sample read-out is not available in the UI while you adjust the DMC adaptive amount? Also, reading from another thread that the "new theory" seems to work with "Fixed rate" sampler type where I don't have the Sampler rate render element to help me understand where rays were distributed in the scene....

[adanmq]

jasonhuang1115

If you use Fixed and 0% adaptive amount like i propose some post before you can make the calculations by "hand", just multiply (Fixed Subdivs² * MSR) to get the "per-pixel" secondary samples . But there is no exact "method" i just find this approach useful to better understand the engine itself and be able to better control it. It depend of the project/scene. And the new additions in Vray 3.1 makes everything simpler.

[jasonhuang1115]

Thanks, adanmq. Which additions in Vray 3.1 are you referring to?

jasonhuang1115

If you use Fixed and 0% adaptive amount like i propose some post before you can make the calculations by "hand", just multiply (Fixed Subdivs² * MSR) to get the "per-pixel" secondary samples . But there is no exact "method" i just find this approach useful to better understand the engine itself and be able to better control it. It depend of the project/scene. And the new additions in Vray 3.1 makes everything simpler.

[adanmq]

Maybe it´s more accurate 3.0 additions, like MSR, the possibility of use a progressive sampler to quick feedback, the optimized BF or the sample hints for example apart from all the "cool" things.

[peterguthrie]

Despite trying several times to find a workflow that suits me, I default back to universal settings on most projects.

adanmq's method sounds promising, but its still too complicated for me..

In general, I'm usually fairly happy with universal, but would LOVE to be able to tell vray to spend a bit less time on areas like trees/grass and more time on things like plain white walls (where noise is undesirable)

Now that we have MSR and even the new vray/properties subdivs multiplier (in the nightlies) there must be some clever 'for-idiots' approach that could be almost as universal as vlados famous universal settings... anyone?