[Diaries - Beginner] Cloud Rendering - Can't Render Fast Enough? Just Put It In The Cloud! (Patreon)

Introduction

If you are an aspiring 3D artist, have you ever wondered, despite every effort, you could never render fast enough to make your 3D career feasible? I mean, not a lot of artists can afford the latest NVIDIA RTX 3000 or RTX 4000 series. As such, the rendering time makes it infeasible for a lot of people to jump on the 3D bandwagon. Fortunately, the era of information dawns in the 21st century. Precisely around the 2010s, several Canadian Scientists managed to crack the secret of Artificial Intelligence. The foundation of AI was created around the 1950s, when a Psychologist in the USA asked the question, "What if we can make a brain inside of a computer and simulate it?" However, this research gets overshadowed by astrophysics, and the time between the 1950s to 2010s is commonly referred to as "The AI Winter".

Because of the tireless effort of the Canadian Scientists in the previous paragraph, the demand and value of data and information skyrocketed. The reason is that Artificial Intelligence feeds on the data that was provided by the internet. Besides that, the demand for computing power also skyrocketed to process Terrabytes of data and information that existed on the internet. And thus, the era of Cloud Computing was here.

So what is Cloud Computing, and how the cloud could help us render our project? 

Cloud computing roughly means lending computing power from the cloud. Then what is The Cloud? The Cloud is a rough definition of dozens of infrastructures on the internet. Usually, these infrastructures are owned by big tech companies, which include Facebook (Meta), Amazon, Apple, Alibaba, Microsoft, IBM, and more. Due to the exponential growth of data, these tech companies usually keep upscaling their infrastructure to accommodate their customers. However, to prevent any energy waste from unused infrastructure, tech companies usually sell their computing service from these unused infrastructures to the public and even their competitors.

This eventually begs the question, "Can we use the infrastructure of the internet to render our project?" The answer is YES! We can use the infrastructure of these companies to render our Blender project.

These are my findings regarding the place where you can render your blender project and where you cannot render your Blender project.

A. Feasible Cloud Platform To Render Your Blender Project

1. Google Colab

Criteria :

• Feasible: YES
• Affordable: YES (Pro => $9-$13/mo depending your region)

Description:

Google Colab is a python-based cloud notebook. Python is a programming language that is used for Data Science, Data Processing, and Data Engineering. It is primarily used to process data and train Artificial Intelligence using GPU or TPU. 

Infrastructure :

Nvidia TESLA P100, TESLA K80, TESLA T4, TESLA V100

Pros :

• Affordable Pro Options (I recommend this)
• Extreme Versatility, limitless way to render your project (render engine, GPU change, input location, output location, etc)
• You can go to blendendertool for an easy way, but your options are limited

Cons :

• 12-hour usage limit for free users, otherwise go to GPU jail
• Requires programming knowledge to fulfill (Pros->B) and troubleshoot coding/rendering problems

Performance Test (I can only testify for this) [600 samples, 2048 tiles]

• NVIDIA GTX 1060 = 600 frames in ~1 week
• Google Colab = 600 frames in ~1 Day

2. Sheep It - sheepit-renderfarm.com 

Criteria :

• Feasible: YES
• Affordable: YES (FREE)

 Description:

Sheep It is a decentralized cloud rendering platform (also known as a rendering farm) where users help each other in rendering their projects. In order for your projects to get rendered, you need to help other users first.

Infrastructure :

Varies (It is decentralized)

Pros :

• The most affordable option for cloud rendering (it's free dude)
• Useful for projects with hundreds or thousands of frames

Cons :

• Your projects get waitlisted until you help other users first
• Can ruin your schedule if you don't have enough computing power to help other users

3. Fox Render Farm

Criteria :

• Feasible: YES
• Affordable: Hourly charge (Good for Image rendering, bad for animation)

Description:

Fox render farm is a cloud rendering platform. They usually charge you per hour of infrastructure usage. Consider using it wisely and don't use it for animation. If you want to render an animation, consider buying GPUs than using this.

Infrastructure :

Lots of them (NVIDIA Tesla V100, RTX 3090, GTX 1080, GTX 750, etc)

Pros :

• Convenient to use

Cons :

• Too Expensive for animation (Just Buy A GPU Bruh)
• Projects need to be properly prepared before rendering (usually texture packing, I don't know if you can choose rendering engines on the fly or even output files)

4. Amazon Web Service (AWS), Microsoft Azure, Alibaba Cloud

Criteria :

• Feasible: YES
• Affordable: NO (~$100+/mo for GPU Supported Machines)

Description:

These are the cloud services of big tech companies. You can do anything on this platform. From hosting a website to building a complex computational model for identifying cancer.

Infrastructure :

Lots of Them (Mostly NVIDIA TESLA)

Pros :

• One of the most advanced cloud computing platforms known to humanity, commonly used by numerous banks and top governments, alongside other popular startups

Cons :

• Too Expensive
• Requires programming knowledge to USE (Not only render)
• Cannot use Blendertool for this

5. Kaggle

Criteria :

• Feasible: YES
• Affordable: YES (FREE)

Description:

Kaggle is a data science website that is aimed at sharing knowledge and resources with other data scientists. It is similar to google Colab. However, you can join Research and Development with other Data Scientists and join data science/AI Development competitions that were usually sponsored by big companies in Kaggle. Some of these companies include Walmart, amazon, google, Facebook, etc. Besides that, Kaggle also attracts top research universities from around the world to conduct collaborative research and competition in Kaggle. These include MIT, Harvard, Stanford, John Hopkins, Oxford, Cambridge, KU Leuven, Karolinska Institute, and more.

Infrastructure :

Mostly NVIDIA Tesla

Pros :

• Free to use
• Good integration with google drive
• you can find some open-source notebooks that you can use to render your project at Kaggle

Cons :

• 30 hours/week GPU Limit
• 6-hour running session cap
• Account limit/tracking (you cannot create an account with an existing name in their database)
• No premium options
• Requires programming knowledge to USE (Not only render)

B. Non Feasible Platform To Render Your Blender Project

1. Paperspace Gradient

Criteria :

• Feasible: NO
• Affordable: NO ($8 upfront, $1.1/hour GPU Usage)

Description:

Paperspace Gradient is similar to Google Colab. It is a python-based notebook designed to train Artificial Intelligence.

Infrastructure:

NVIDIA Quadro P4000, RTX 4000, RTX 5000, etc

Pros :

• Convenient to use

Cons :

• Too Expensive
• You'll get banned if you try to install anything lmao (first-hand experience)
• Requires programming language to USE (Not only render)

Closing Note :

Don't combine cloud rendering with local rendering when rendering a scene. For some reason, there is a noticeable quality difference even though the rendering engine is the same.