ANIMATION

تسریع سرعت رندر در Vray

V-RAY RENDER OPTIMIZATION

این آموزش یکی از دقیق ترین آموزش هایی است که تا به امروز در مورد بهینه کردن تنظیمات رندر تهیه شده است 

مترجم: لفظ Optimization از نظر لغوی به معنای بهینه سازی کردن اما در این بخش برای اینک شما خواننده عزیز درک بهتری از مقاله داشته باشید باید کمی ترجه تحت لفظی این م کلمه را عوض کرد.optimization، هرگونه مدیریتی که بروی صحنه خود داشته باشیم میتونه در قالب این لغت بگنجه اما مدیریت در راستای افزایش سرعت رندر، این موضوع از مدیریت صحنه و استفاده کردن از  دستوراتی مثل vray Proxy گرفته تا درک درسته Subdivision در متریال و یا حتی نحوه کپی کردن مدل ها در صحنه.پس در کل به مدریت حالا چه در صحنه چه در تنظیمات رندر میتوانیم بگویم Optimization  

v1.3

 - AUGUST 2014

Render scene courtesy of Chaos Group
 

مقدمه :

این در مقاله تلاش شده روش های به شما آموزش داده بشود که با مدیریت کردن تنظیمات رندر خود کمترین زمان رندر و بیشترین کیفیت رندر را داشته باشید.

اغلب برای کاربر سردرگمی هایی اساسی وجود داره که چه تنظیمات رندری میتونه یک تنظیمات رندر ایده آل باشه .در اکثر زمان های میبینیم که کاربران تنظیمات رایجی که در فضای مجازی ببه انتشار گذاشته شده را استفاده میکنند  و فکر میکنند که در صورت بهره گیری از این تنظیمات زمان رندر آنها سریعتر و کیفیت رندر آنها بهتر میشود که مطمنا سخت در اشتباه هستند ، الزاما هر تنظیماتی که ما میبینیم نمیتواند خوب باشد، اما با کمی درک  بهتر عملکرد تنظیمات رندر VRay شما میتوانید به یک تنظیمات خوب با یک سرعت رندر بالا و کیفیت عالی دست پیدا کنید بدون اغراق میتوانم بگویم شاید 3 تا 13 برابر سریعتر از تنظیماتی که بصورت عمومی کاربر های مختلف تا به امروز منتشر کرده اند .

خوب ما در ابتدا باید فلسفه وجودی بعضی از بخش های vray  رو تعریف کنیدم مثلا اینکه Ray Tracing  و یا Vray  Sampling چه عملکردی ارند و بعد از اون با تست چند صحنه به شما نشان میدههم که دقیقا چطور یک رندر را میتوانید optimize کنید و زمان رندر کوتاه و کیفیت فوق العاده ایی داشته باشیدو در نهایت من قدم به قدم روش optimize کردن  صحنه ها را به شما آموزش میدهم

 

RAYTRACING 101

 

زمانی که یک رندر شروع میشود، ابتدا Rays ( اشعه ها ) از رندر ما به صحنه فرستاده میشود تمامیه اطلاعات مربوط به مدل های موجود در صحنه را جمع آوری و آنالیز میکند و در اخرین مرحله رندر محو میشود.منشا تمام این Ray از دوربین میباشد که به آنها گفته میشود Primary Rays (اشعه ها  اصلی ) البته اصطلاحات دیگری هم برای این اشعه ها بکار میرود همچون (Camera Rays یا Eye Rays) که این اشعه ها توسط Vray Image Sampler (که ما با لفظ Anti-Aliasing با AA میشناسیم ) کنترل میشود 

هنگامیکه Primary Ray  برخورد میکنند با مدل های موجود در صحنه به تعداد بیشتری تقسم میشوند ، Ray جدید بوجود اماده به سمت مناطق دیگر صحنه حرکت میکننید و این تقسیم و برخوردها آنقدر اتفاق می افتد تا تمامیه اطلاعات صحنه از جمله  Lighting, Shadows, Global Illumination (GI), Reflection, Refraction, Sub-surface Scattering (SSS) را بطور کامل جمع آوری کند . به Ray  که بعد از برخورد اول بوجود میآید Secondary Ray  گفته میشود که توسط Vray DMC Sampler  کنترل میشود .

 


We'll first cover some of the underlying concepts behind how ray tracing and V-Ray's sampling works. Then we'll go over an example scene to demonstrate exactly how a render can be optimized to be faster and cleaner. Then we'll learn how to identify the different sources of noise a scene can have. And finally I'll provide a step-by-step procedure to optimize any scene to render with an ideal balance of quality and speed.

FIGURE 01

تصویر بالا یک دیاگرام ساده از  Raytracing ( برخورد اشعه ها با سطوح ) میباشد 

Primary Rays از دوربین متساعد میشوند به سطوح مدل های موجود در صحنه برخورد میکنند و سپس به تعداد بیشتری تبدیل میشوند (ُSecondary Rays )  و بدین شکل از کل صحنه نمونه برداری میکنند 

از اینجای مقاله به بعد هرجای مقاله که اسم Sample آورده شد منظور ما Ray  میباشد بخاطر اینکه تمام اهداف Ray  نمونه برداری ( Sample ) و جمع آوری اطلاعات موجود در صحنه میباشد پس اساسا باید بگوییم که  Rays = Samples

اکر به تصویر بخوبی نگاه کنید کاملا نشان میدهد که چه اتفاقی در صحنه رخ میدهد و میبینید نیاز است که Primary and Secondary Samples بیشتری برای جمع آوری اطلاعات استفاده کنیم.هرچقدر میزان نمونه برداری و جمع آوری اطلاعات از یک صحنه بیشتر شود کیفیت رندر بسیار بهتری و در نتیجه Noise  کمتری در رندر خواهیم داشت.پس حالا میتوانید گه وجود Noise ها در یک رندر به چه دلیل است، بله درسته به دلیل کبود اطلاعات و یا ناقص بودن اطلاعات از آن صحنه است.Noise های بوجود آمده در یک صحنه به این معناست که Vray توانایی جمع آوری اطلاعات از آن صحنه را نداشته و نمیداند که در آن صحنه چه اتفاقاتی افتاده است بنا بر این به منظور کاهش Noise شما نیاز دارید که اطلاعات بیشتر یا اینطور بگویی نمونه برداری یا همان Sampling  بیشتری در صحنه داشته باشیم.

میزان Parimary Sample  فرستاده شده به صحنه توسط Min Subdivs, Max Subdivs و  Color Threshold که در بخش Image Sampler وجود دارند کنترل میشود و میزان Secondary Sample که در صحنه وجود میآیند توسط تنظیمات Subdivs موجود در Lights / Global Illumination / Materials و تنظیماتن Noise Threshold واقع در DMC Sampler کنترل میشود .(در Vray Maya لفظ Noise Threshold با کامه Adaptive Threshold  شناخته میشود )

 

In order to accurately figure out what's going on in a scene, many Primary and Secondary Samples are needed to be taken. 


So to recap the important terms:


Ray =

Sample

Primary Samples =

کنترل میزان Sample ها توسط Vray Image Sampler یا همان  (Anti-Aliasing, AA) که البته هر صحنه با توجه به  Motion Blur و Geometry, Textures, Depth of Field مخصوص به خوش نمونه برداری میشود .

Secondary Samples =

Sample این بخش توسط Vray DMC Sampler  کنترل میشود که  این بخش هم با توجه به  Lighting, Global Illumination (GI), Shadows, Material Reflection & Refraction و  Sub-Surface Scattering (SSS( مربوط به خودش نمونه برداری میشود.

Noise =

 به دلیل کمبود اطلاعات.

Subdivs =

میزان اعداد در این بخش توان 2 دارند.  Subdivs2 = Samples.
برای مثال:  Subdivs  8 = 64 Samples. 82 = 64 

در ادامه این مقاله شما یاد میگیرید که چطور با استفاده درست از Primary و Secondary Sample  رندر باکیفیت تری با کمترین میزان Noise  و کمترین زمان رندر داشته باشد

 SampleRate RENDER ELEMENT چیست:

SampleRate render element یکی از مهمترین ابزار هایی است که کمک میکند رندر بهترین Optimize را داشته باشد. Vray از این طریق به ما نشان میدهد که  دقیقا  (Image Sampler (AA بروی هر پیکسل چگونه آنالیز شده است.این کار با اختصاص یک رنگ خاص برای هر پیکسل  با توجه به میزان  (Primary Sample (AA در  هر رندر مشخص میشود 

  • رنگ آبی کمترین میزان (Primary Sample (AA بروی هر پیکسل
  • رنگ سبز مقدار متوسطی از (Primary Sample (AA بروی هر پیکسل
  • رنگ قرمز بیشتر میزان  (Primary Sample (AA بروی هر پیکسل

 

FIGURE 02

SampleRate render element که در سمت راست قرار دارد به ما نشان داده که چه مقدار Primary Sample برای هر پیکسل اختصاص داده شده است

بنابر این وقتی که برای یک صحنه برای   image Sampler (AA)  1min , 10max Subdivs  ست میکنیم  به این معنیست که  1min and 100max Primary Samples
  •  رنگ آبی یعنی 1 Primary Sample
  • رنگ سبز یعنی 50 Primary Sample
  • رنگ قرمز  یعنی 100 Primary Sample 
 

بنابر این وقتی که برای یک صحنه برای   image Sampler (AA)  1min , 100max Subdivs  ست میکنیم  به این معنیست که  1min and 10000max Primary Samples

  •  رنگ آبی یعنی 1 Primary Sample
  • رنگ سبز یعنی 5000 Primary Sample
  • رنگ قرمز  یعنی 10000 Primary Sample 
 

 

  • A Red pixel means that 10000 Primary Samples were taken.

 

نمونه صحنه آزمایشی ، برای درک بهتره نحوه عنملکرد Vray

برای این آموزش ، ما بروی یک صحنه ساده که شامل یک صفحه چند کره که داره متریال های مختلفی هستند کار میکنیم، این متریال ها دارای  (including diffuse, glossy reflection, glossy refraction, and SSS) و نور های صحنه که :2 تا Area Light  و یک Dome Light با یک HDRI میباشد.Global Illumination  روشن میباشد و موتور های محاسبه گر هم بروی Brute Force + Light Cache تنظیم شده است.این صحنه را ازاز لینک روبروی دانلود کنید. دانلود 

تنظیمات ابتدایی این صحنه :


  • Image Sampler (AA) set to 1min & 8max Subdivs.
  • Lights, GI, and Materials all set to the default of 8 Subdivs.
  • Noise Threshold set to the default of 0.01.
  • All other render settings are left at their defaults.

FIGURE 03

Baseline Render
1min & 8max Subdivs = Image Sampler (AA)
8 Subdivs = Lights, GI, and Materials each.

 

Now lets go over exactly what's happening in this baseline render. Through the render settings, you're telling V-Ray:

"I'm allowing you to use up to 64 (8 Subdivs) Primary Samples (AA) per pixel to figure out what's going on in this scene and reduce the noise as close as you can to my specified noise threshold... BUT for each of those Primary Samples you take, you're only allowed to take 1 additional Secondary Sample to figure out what's going on for each Light, GI, and Material."

At this point you may be asking:

"Wait, only 1 Secondary Sample for the Lights, GI, and Materials each? Shouldn't it be 64 Samples (8 Subdivs) as we've set them?"

Well it's important to note that even though the Lights, GI, and Materials are set to 64 Samples (8 Subdivs) each - V-Ray internally divides this value by the AA Max Samples value of your scene. So instead of the 64 Samples for the light and material each as you might expect, this gets divided by the AA Max of 64 Samples (8 Subdivs), which results in only 1 Secondary Sample being taken for the Lights, GI, and Materials each. (64 Secondary Samples / 64 Primary Samples = 1 Secondary Sample).


The reason V-Ray does this is because it's internal formulas are set up in a way that attempts to automatically balance it's two samplers. The thinking behind this is that with more Primary Samples being taken of your scene, a proportionally smaller amount of Secondary Samples are needed to figure out exactly what's happening in the scene (which we'll soon learn, this isn't always desirable). This balancing of the Image Sampler and the DMC Sampler can be a bit confusing at first, but the important thing to take away is this: Whenever you increase your Image Sampler (AA) settings, V-Ray tries to compensate by internally decreasing your DMC Sampler settings. Later on if you're interested in learning more about how V-Rays internal formulas work - you can check out the DMC Calculator I've coded - but for right now it's not necessary.



So lets get back to the rendering:

V-Ray finishes the render as best as it can manage, but warns you (by the many RED pixels in the SampleRate render element):

"I was NOT able to figure out what's going on in all of this scene according to the level of quality (noise threshold) you want it to be at! - A lot of the time I had to use up all of the 64 Primary Samples with 1 Secondary Sample per Light, GI, and Material you allowed me to use for each pixel, but it still didn't provide me with enough information in those areas."

If we take a look at the render - we'll notice that while the geometric detail (edges of objects) seem fairly clean and defined, there are indeed noisy areas of the image - specifically noticeable in the reflections and shadows. So we've got this noisy baseline render, and we have two options to reduce the noise to meet our desired level of quality (noise threshold):


Option #1 =

Increase AA Max Subdivs - Let V-Ray take more Primary Samples to figure out the scene - but again only 1 additional Secondary Sample for the Lights / GI / Materials each per Primary Sample.

Option #2 =

Increase Lights / GI / Materials Subdivs - Tell V-Ray to stay at the same amount of Primary Samples - but instead allow it take more Secondary Samples with each Primary Sample to better figure out the scene.

EXAMPLE SCENE - OPTION #1 - INCREASED AA MAX SUBDIVS

So lets first try what most people usually do to get a high quality (low noise) render - adopt the so-called 'Universal V-Ray Settings' and let V-Ray take as many Primary Samples (AA) as needed to eliminate the noise.


  • We'll increase the Image Sampler (AA) to 1min & 100max Subdivs.
  • We'll leave the Lights, GI, and Materials set to 8 Subdivs each.
  • We'll reduce the Noise Threshold to 0.005 to tell V-Ray we really want a noise-free render.

FIGURE 04

Option #1 Render - Increased AA Max Subdivs
1min & 100max Subdivs = Image Sampler (AA)
8 Subdivs = Lights, GI, and Materials each.
0.005 = Noise Threshold.

 

Now lets go over exactly what's happening in this Option #1 render. Through the render settings, you're telling V-Ray:

"I'm allowing you to use up to 10000 (100 subdivs) Primary Samples (AA) per pixel to figure out what's going on in this scene and reduce the noise as close as you can to my specified noise threshold... BUT for each of those Primary Samples you take, you're only allowed to take 1 additional Secondary Sample to figure out what's going on with the Lights, GI, and Materials each."

Again, remember that even though the Lights, GI, and Materials are set to 64 Samples (8 Subdivs) each - V-Ray internally divides these values by the AA Max Samples value of your scene. So instead of 64 Samples, this gets divided by the AA Max of 10000 Samples (100 Subdivs), which results in the minimum of only 1 Secondary Sample being taken for the Lights, GI, and Materials each. (64 Secondary Samples / 10000 Primary Samples = 1 Secondary Sample).


V-Ray finishes the render as best as it can manage, and tells you (by the now mostly BLUE SampleRate render element):

"I was able to figure out all of what's going on in this scene to the level of quality (noise threshold) you want it to be at! - In fact, I was able to figure it out well before I had to use all 10000 Primary Samples with 1 Secondary Samples per Lights, GI, and Materials you allowed me for each pixel."

We take a look at the Option #1 render and see the noise has definitely been improved compared to the baseline render. The render time has increased by 11min 44s (9.8x longer) compared to the Baseline Render, but that's to be expected with a higher quality render, right? At this point, most people would think this is as good as it gets, and call the render FINISHED!


...But what happens if we were to try that Option #2 we discussed earlier? Instead of increasing the AA Max Subdivs, what happens if we instead opted to only increase the Lights / GI / Materials Subdivs? Well lets find out...

EXAMPLE SCENE - OPTION #2 - INCREASED LIGHT / GI / MATERIALS SUBDIVS

This time we'll try something a little different - we'll tell V-Ray to take same amount of Primary Samples that we originally did in the baseline render - but instead allow V-Ray to take more Secondary Samples with each Primary Sample to better figure out the scene.


  • We'll leave the Image Sampler (AA) set to the original baseline render's settings of 1min & 8max Subdivs.
  • We'll increase the Lights, GI, and Materials to 80 Subdivs each.
  • We'll leave the Noise Threshold set to the original baseline render's default of 0.01

FIGURE 05

Option #2 Render - Increased Lights, GI, and Materials Subdivs
1min & 8max Subdivs = Image Sampler (AA)
80 Subdivs = Lights, GI, and Materials each.
0.01 = Noise Threshold.

 

Once more, lets go over exactly what's happening in this Option #2 render. Through the render settings, you're telling V-Ray:

"I'm allowing you to use up to 64 (8 subdivs) Primary Samples (AA) per pixel to figure out what's going on in this scene and reduce the noise as close as you can to my specified noise threshold... AND for each of those Primary Samples you take, you're allowed to take up to 100 additional Secondary Samples to figure out what's going on with the Lights, GI, and Materials each."

Again, remember that even though the Lights, GI, and Materials are set to 6400 Samples (80 Subdivs) each - V-Ray automatically divides these values by the AA Max Samples value of your scene. So instead of 6400 Samples, this gets divided by the AA Max of 64 Samples (8 Subdivs), which results in only 100 Secondary Samples being taken for the Lights, GI, and Materials each. (6400 Secondary Samples / 64 Primary Samples = 100 Secondary Sample).


V-Ray finishes the render as best as it can manage, and tells you (through the SampleRate render element):

"I was able to figure out almost all of what's going on in this scene to the level of quality (noise threshold) you want it to be at! - In fact, most of the time I was able to figure it out well before I had to use all 64 Primary Samples per pixel! All those extra 100 Secondary Samples per Light, GI, and Material provided each Primary Sample with so much more information this time!"

We take a look at the Option #2 render and see the noise has definitely been improved compared to the baseline render. The render time has increased by 4m 38s (4.5x Longer) compared to the Baseline Render, but that's to be expected with a higher quality render.


But here's where things start to get interesting...

When we compare Option #2's render against Option #1's render, we can see that Option #2 gave us a cleaner render!

 

And look at that! - Option #2 finished 2.2x faster than Option #1! - 5m 58s for Option #2 versus 13m 04s for Option #1!

FIGURE 06

Option #1 Render on the left, and Option #2 Render on the right.
At the bottom, the renders zoomed 400% to better show noise levels.

 

Why is this? Why did increasing the DMC Sampler settings (Lights / GI / Materials Subdivs) rather than increasing the Image Sampler (AA) settings result in a cleaner AND faster render for this scene? We even set Option #1 to have a lower noise threshold, yet it still turned out noisier than option #2! Well the answer lies in what we originally noticed about the Baseline render...

HOW OPTIMIZATION WORKS

In our Baseline Render, we saw that while the edges of objects looked clean and defined, the noise seemed to reside primarily in the reflections and shadows. Well, if you remember what we learned earlier: Primary Samples (AA) specialize in figuring out the geometry, textures, depth of field, and motion blur in a scene. While Secondary Samples specialize in figuring out lighting, GI, shadows, materials, etc.


So in the case of fixing the noise in the Baseline render, the choice between Option #1 and Option #2 is actually a no-brainer! Why use a screwdriver to do a hammer's job? The Image Sampler (AA) had already done what it was designed to do - make the geometric detail (edges of objects) clean and defined. So instead of firing a bunch of additional Primary Samples (AA) at the scene to clean up noise - it's better to allocate those additional samples to the DMC Sampler (Lights / GI / Materials Subdivs) so it can properly do what it was designed to do - clean up the noise in the lighting, shadows, GI, reflections, and refractions. There's our answer!

 

And now can we begin to understand why the 'Universal V-Ray Settings' of 1min and 100max AA is generally not going to be the most efficient method to render a scene - in fact it never was meant to be the most efficient method! The Universal V-Ray Settings were designed to make V-Ray accessible and easy for users who don't care about render optimization or learning how V-Ray works under the hood. It's simply a method to put V-Ray on auto-pilot. It allows a user to control all render quality by adjusting only one setting - the noise threshold. If there's too much noise in a render, just lower the noise threshold, and V-Ray will keep firing Primary Samples (AA) at the scene until it eventually reaches the noise threshold - guaranteeing a nice looking render with minimal understanding of how V-Ray works. But that render is generally not going to be as clean or render as fast as if you take the time to understand how V-Ray works and balance the Image Sampler and DMC Sampler according to a scene's demands!


And just to really drive the point home - the Option #2 Render can still be optimized even further! Using some additional tricks listed in the procedures at the end of this tutorial, we can reduce it's render time from 5m 58s to 4m 53s with only a slight hit in noise amounts! For a final render speed increase of 2.7 times faster than the Option #1 Render!

FIGURE 07

Option #1 Render on the left, and Option #2 Render optimized even further on the right. For a total render speed increase of 2.7x!

 

Here's another example of optimization, this time with a more production oriented scene...

The optimized render (right) renders nearly 35% faster than the universal settings render (left) while reducing noise and improving render quality. Also note how the reflections have become more accurate - noticeable on the floor towards the end of the hallway.

FIGURE 08

'Universal V-Ray Settings' on the left, and the Optimized render on the right.
Render scene courtesy of Peter Guthrie

 

IDENTIFYING SOURCES OF NOISE

The key to properly optimizing a render is to correctly identify which aspects of a scene are causing noise, and assigning the right sampler with enough samples to attack that noise at it's source. Some scenes will require more samples for the Image Sampler, while others (like the ones shown in the examples above) will require more samples for the DMC Sampler. As a general guideline:


Circumstances where the Image Sampler (AA) will require larger amounts of Primary Samples to eliminate noise:

  • Fine geometric detail like Hair, Grass, Foliage, etc.
  • Very fine texture detail like weaves, tiny bump map details, etc.
  • Scenes with shallow Depth of Field or heavy Motion Blur.

Circumstances where the DMC Sampler will require larger amounts of Secondary Samples to eliminate noise:

  • Large light sources that cast soft shadows.
  • Materials with strong glossy Reflection or Refraction.
  • Scenes with prominent Global Illumination - particularly indoor scenes.

Noise caused by the Image Sampler (AA) is luckily very easy to spot to the naked eye. It manifests itself in jagged or unclear object edges, undefined texture detail or effects like Moiré patterns, and grainy depth of field or motion blur.


Noise caused by the DMC Sampler can be a bit trickier to see exactly what's causing it. Luckily we have some handy tools at our disposal to help us figure it out - V-Ray's Render Elements: Lighting, Global Illumination, Specular, Reflection, and Refraction. By looking through these various render elements, you can quickly isolate and check the levels of noise caused by any of these individual aspects of your scene.

FIGURE 09

By looking at the Reflection render element, we can see the amount of noise caused by material reflections alone.

 

OPTIMIZATION PROCEDURE

So now that we have a solid understanding of what can be gained by properly utilizing V-Ray's samplers, and how to identify the sources of noise in our scene, we can follow a step-by-step procedure to help optimize whatever scene you may be dealing with.


The general idea behind this procedure is to begin by setting the Image Sampler (AA) Max just high enough to clearly define the Geometry, Textures, Depth of Field, and Motion Blur detail in your scene, and then tune each of the DMC Sampler's aspects (Lights/Specular/Shadows, GI, Material Reflection/Refraction) one at a time to eliminate noise. So we'll start with default V-Ray settings, and bring our render to a very basic state to start the optimization process:


    1. In V-Ray's Global Settings rollout - uncheck 'Shadows', 'Reflection/Refraction', and 'Glossy Effects' to disable them.
    2. In V-Ray's GI (Indirect Illumination) tab - disable Global Illumination.
    3. In V-Ray's Render Elements tab - add: SampleRate, Lighting, Global Illumination, Specular, Reflection, Refraction.
    4. Optional: Reduce your render frame size for faster feedback during the optimization process.
      At this point, your render should be nothing but a very plainly lit diffuse render of your scene. [Example]

Tuning the Image Sampler (AA):

    1. Gradually raise the Image Sampler (AA) Max Subdivs value until all geometry edges, texture detail, depth of field, and motion blur in your scene appears crisp and clear. Check the SampleRate Render Element for red areas and reduce them as needed by increasing your Image Sampler (AA) Subdivs Max value until they only reside along the very edges of your object and texture details. Once these details appear nice and crisp, leave your Image Sampler (AA) settings alone.

Tuning the scene's Lighting / Shadows / Speculars:

    1. In V-Ray's Global Settings rollout - Check 'Shadows' to re-enable them.
      At this point your render should have lighting and shadows, but still diffuse materials. [Example]
      Notice that noise (along with more red pixels in the SamplerRate render element) has been re-introduced to the render from enabling your scene's Lighting.
    2. Raise the Subdivs value in your Lights until the noise in the shadows and lighting become acceptably smooth and clear. It's easiest to see the effect of raising these Subdivs values by checking your Lighting render element. As you raise these Subdiv values, the red pixels in your SampleRate render element should again be receding back to the very edges of your objects and textures while most other areas gradually become more blue or green - indicating that less Image Samples (AA) are being needed. 

      A note about Speculars: The quality of specular highlights in your scene are also controlled by your Lights Subdivs value. As a general rule of thumb, once the shadows of a light are free of noise, the specular highlight for that light should also be free of noise. But for now we wont be able to tell since we currently have Glossy Effects (which Specular Highlights are a part of) globally disabled for this stage of the optimization process. So later on when wrapping up the optimization process if you notice any noisy specular highlights in the Specular render element, increasing the light's Subdivs value is how to fix it.

Tuning the scene's Global Illumination (GI):

    1. In V-Ray's GI (Indirect Illumination) tab - re-enable Global Illumination.
      I recommend using Brute Force for the primary engine, and Light Cache for the secondary engine unless dealing with specific cases (like a camera fly-through) where using a cached GI method like Irradience Map is more appropriate.
      At this point your render should have normal lighting, shadows, and now GI, but still diffuse materials. [Example]
      Again notice that noise (along with more red pixels in the SamplerRate render element) has been re-introduced to the render from enabling your scene's GI.
    2. Raise the Subdivs value of your Brute Force GI until any noise in the Global Illumination becomes acceptably smooth and clear - this time using the Global Illumination render element pass to check the noise levels. Again, as you raise the Subdivs value, the red pixels in your SampleRate render element should be receding back to the very edges of your objects and textures while most other areas gradually become more blue or green - indicating that less Image Samples (AA) are being needed. 

      A note about Hair: If using the VRayHairMtl in your scene, the quality of your hair's reflection and refraction is in fact controlled by your scene's Global Illumination Subdivs value. You can verify this by noticing how hair's reflection and refraction doesn't appear at all in the Reflection or Refraction render elements until GI is turned on. This is because the VRayHairMtl's reflection and refraction components are entirely dependent on GI, and this is also why the VRayHairMtl has no separate Subdivs values for it's reflection and refraction. So now is also the time to be tuning your hair reflection and refraction quality.

Tuning the scene's Reflections and Refractions:

  1. In V-Ray's Global Settings rollout - Check 'Reflection/Refraction' and 'Glossy Effects' to re-enabled them.
    At this point your scene should render as normal, with lighting, GI, and material reflection & refraction. [Example]
    But again, notice that noise (along with more red pixels in the SamplerRate render element) have been re-introduced to the render from enabling your scene's Reflections & Refractions.
  2. Raise the Subdivs values in your materials' Reflections until any noise in them becomes acceptably smooth and clear. It's easiest to see the effect of raising these Subdivs values by checking your Reflection render element pass. Again, as you raise these Subdiv values, the red pixels in your SampleRate render element should again be receding to the very edges of your objects and textures while other pixels gradually become more blue or green - indicating that less Image Samples (AA) are being needed.
  3. Repeat the previous step, this time with your scene materials' Refractions.

Your render should now be optimized for an ideal render time according to the level of quality (noise threshold) you have it set at!

If you desire less noise, lower the Image Sampler's 'Color Threshold' value, and raise Lights / GI / Materials Subdivs as needed.

FIGURE 10

An animated GIF illustrating how allocating more samples to the DMC Sampler reduces the stress on the Image Sampler. The Image Sampler's Min & Max Subdivs stay the same, while the scene's Lights / GI / Materials Subdivs rise - resulting in less Primary Samples (AA) being needed while quality improves.

 

ADDITIONAL TIPS & TRICKS

  • In the examples in this tutorial, I kept the Lights / GI / Materials all at the same Subdivs values for the sake of simplicity and to make the concepts easy to understand. But it's important to remember that a properly optimized scene will have varying Subdiv values specifically tailored to the needs of each of these secondary aspects of the scene.
    For example: A material that's only 5% reflective probably wont need too many reflection samples to prevent noise from being visible in the final image, since it's final RGB value will be 95% dependent on the material's diffuse or refraction components. But if that same material was instead 95% reflective, it will need more reflection samples to prevent noise being visible in the final render, since the reflection will now be the main contributing factor of the material's final RGB values. The same goes for material glossiness - the more glossy a material's reflection or refraction is, the more samples it'll need to eliminate noise. The same applies for lights - the larger the light, the softer the shadows, and the more samples it'll need eliminate noise.
  • Keep the DMC Sampler's Adaptive Amount setting below 1.0. There are many instances where setting this value at or near the maximum value of 1.0 can give really unpredictable results. The default value of 0.85 is a good place to leave it set.
  • Experiment with unlinking the Image Sampler's Color Threshold value from the DMC Sampler's Noise Threshold by unchecking the 'Use DMC Sampler Threshold' checkbox under V-Ray's Image Sampler rollout. Then increase the Color Threshold value in small increments to help the Image Sampler not confuse noise caused by Lights / Shadows / GI / Materials for geometric or textural detail. Raising this value can potentially re-introduce noise in your image, so you can either raise your secondary Subdivs values higher to compensate, or just accept the slightly noiser but faster render times. It takes a bit of experimentation to see if it's worth it for your particular scene.
  • Experiment with disabling the Image Filter of the Image Sampler if you don't have very small repeating detail patterns in your scene. V-Ray's Image Sampler (AA) already does a good job resolving and smoothing detail - so additional filtering sometimes isn't needed.
  • Experiment with disabling the Filter Maps checkbox in V-Ray's Global Settings rollout, or lowering the filter/blur values of individual bitmaps used in your scene's materials to values like 0.1 or 0.2. Especially consider disabling the filtering of opacity maps for materials like leaves in trees. Disabling or reducing the filter/blur values of maps will have a definite render speed increase while making your maps appear sharper, but can also have the side effects of increased RAM usage, and possible flickering or scintillation of maps in animations.
  • If using Light Cache for your GI's secondary bounces - enable 'Use for Glossy Rays' and 'Retrace Threshold' to help speed up the calculation of glossy materials.
  • If using V-Ray 3.0 - try enabling the Embree raycaster via the 'Use Embree' checkbox in the V-Ray's System rollout. In many scenes the Embree raycaster can yield a great reduction in render time with no visible difference in quality.
  • Reduce V-Ray's Bucket Size to something like 32x32 or 24x24 to avoid larger individual buckets getting stuck on a particularly difficult part of the render while other the buckets finish rendering and leave the image - causing your CPU to only use a fraction of it's available power to finish the last bit of the render. V-Ray 3.0 dynamically reduces Bucket Size towards the end of a render to help avoid this for you.

CONCLUSION

In closing - it's important to keep in mind that every scene is different, and therefore has different needs from V-Ray's two samplers. Settings that optimize one scene can potentially grind another scene to a halt - so please remember: Your settings and ability to optimize a render can vary greatly from one scene to the next. With a bit of trial & error, practice, and patience, you'll gain the experience to intuitively know what settings a scene calls for.


So hopefully that helped to clear up the often confusing topic of V-Ray Render Optimization for you! I wish you fast renders and the best of luck with your projects!


Special thanks to:

Toni Bratinevic and John O'Connell for the information they've generously shared with the community on the subject. Peter Guthrie for lending his gallery scene for use in the examples. John Rouse for proofreading and testing the procedures in this article.

نوشتن دیدگاه


تصویر امنیتی
تصویر امنیتی جدید

سایتهای مرتبط

UP Studio
استودیو پرزانته معماری
 
 

Pooria Abbasi 
آموزش مدلسازی پارامتریک و احجام پیچیده معماری
 

 

درباره سایت :

سایت حاضر در تلاش است به منظور کمک به افراد علاقمند به شاخه پرزانت حرفه ای معماری تا علاوه بر یادگیری از طریق آموزش های ارائه شده از خبر های مربوطه نیز اگاه شوند.

 

FOLLOW ME

          

S5 Box

Tab 9

dsaaaaaaaaaaaaaaaaa

طراحی سایت طراحی سایت بهینه سازی سایت طراحی اپلیکیشن