The future of rendering
In the past couple of years rendering has been transformed. We investigate the new technologies that have revolutionised the process for both studios and standalone artists
We investigate the new technologies that have revolutionised the process for both studios and standalone artists
“Being able to utilise a large public cloud meant that there would never Be a limit to machine availability” Todd Prives, Google
Rendering has always been the ‘problem part’ in any CGI project. It has been the single biggest cause of headaches, lost nights and panic for artists. Render machines fall over, animation frames can have an unforeseen issue, or there simply isn’t enough time to render to the quality level that is needed without a huge investment in render machines. But with new technology looking to fix these problems, could the way we think of rendering change?
Heads in the cloud
Many of the most popular render engines are Cpu-based, which means that for every new render node, a new PC has to be purchased. This is coupled with a new render license for every new machine. While there has been a recent heating up in the CPU wars with the introduction of AMD Ryzen’s chipset (which has brought prices down), a new PC is still a significant outlay.
With the rise of broadband speed, cloud rendering has become viable for more artists. Cloud rendering enables any artist to hire as many PCS as they need via an online service. These cloud render farms come in various sizes, with vendors offering access to hundreds of machines, which means renders in minutes, as long as there isn’t a queue of other users.
Recently, Amazon and Google have waded into the cloud render space. Amazon offers the ability to create custom render nodes through its EC2 system. Google has taken the direct approach with its 2014 acquisition of cloud render provider, ZYNC Render, which was originally developed as an in-house solution in 2011 for Zero VFX.
Todd Prives, a founder of ZYNC Render who now oversees the continuation and expansion of ZYNC Render as a product manager within Google, explains more about how ZYNC Render was formed: “We had limited financial resources and knew that being able to leverage cloud to accommodate the ebb and flow
of project demands without a massive upfront capital investment would be critical. We wanted to build an integrated solution that automated asset management and synchronisation and gave us direct control and visibility over the render nodes in real time. Being able to utilise a large public cloud meant that there would never be a limit to machine availability, nor would we be forced to wait behind other customers.”
ZYNC Render is able to offer up to 500 machines to every user when required, due to the fact that it can use Google’s vast infrastructure of custom-built computers, which are otherwise engaged in solving web searches or processing selfies.
If a company like Google is involved in the cloud render space, though, how do other companies hope to compete?
Yellowdog is a relatively new British cloud rendering service taking on this challenge. Its solution takes the democratisation of cloud rendering to a new level, as Gareth Williams of Yellowdog explains: “Yellowdog is different from other render farms and cloud companies in that we aren’t limited by a single data centre. The Yellowdog platform provides access to computers from multiple cloud providers, which means animators and 3D artists can access limitless computers.”
Some of these cloud providers are 3D artists themselves, and Yellowdog offers artists the chance to make money from their workstations, to become part of Yellowdog’s cloud solution.
Graphics cards – not just for Gaming
The second major breakthrough has been in the development of new GPU rendering software that shifts the majority of the render workload from the CPU to the computer’s graphics card. This has seen exponential increases in render speed and functionality for many artists. Rendering with GPUS requires a much lower investment than CPU render solutions, as many PCS can have multiple graphics cards installed. With the arrival of super-fast I/O protocols such as Thunderbolt 3, which allows graphics cards to be installed externally, even a suitable Mac or Windows laptop can be an effective GPU render machine.
Many of the original GPU render engines were unbiased, which meant they strictly obeyed the laws of physics, unlike biased solutions such as V-ray, which can be optimised to suit the scene.
Potentially easier to learn than biased engines, there had only ever been a few unbiased solutions based on CPUS, such as Next Limit’s Maxwell Render (which now boasts GPU acceleration as well). This was mainly because comparative render times were slower than biased render engines, although interactive render previews did go some way to help mitigate that.
GPU unbiased engines, such as OTOY’S Octane and Indigo Renderer, have transformed unbiased rendering by dramatically improving render times while keeping the essentials of interactive render previews, which are also accelerated. Nicholas Chapman, managing director of Glare Technologies, who makes Indigo, explains: “Indigo itself was started back in the late 2000s, based on code and research that I was doing on Monte Carlo path tracers. I was fascinated by the way such realistic images could be generated with such relatively simple algorithms.
“In 2011, with Indigo 3.0, we added support for offloading raytracing to the GPU via CUDA and Opencl. It became clear to us that a full GPU rendering solution (materials, raytracing, lighting etc, all running on the GPU) was needed to exploit the performance available with GPUS. We started to work on that, and now it’s available with the release of Indigo 4.0.
“This development has been worth it as Indigo now renders approximately ten times faster on the