case study: territory studio on the benefits of an internal farm
Territory Studio is famous for ITS ground-breaking work in ui, graphics and high-end cgi for Some of The largest box office Successes of recent years
Recent notable work has included Ghost in the Shell, The Martian, and Mission Impossible – Rogue Nation.
Working on such high-profile projects has implications at final render time, which Territory Studio’s creative director, David Sheldon-hicks, explains: “Most of our film projects, including
Ghost in the Shell, are not allowed to go to the cloud for security reasons. Up until very recently security protocol with feature films forbid cloud usage. This is being loosened to some extent, through better encryption and security measures at specialist facilities.”
Also because lead VFX facilities on a film have started to distribute assets to other studios to minimise the translation and rebuilding effort, this has led to other issues, as David explains:
“The Iron Man models we received from ILM for the first Avengers wouldn’t even load into the memory on a graphics card due to the sheer detail of the models. This is often true for film assets.”
This has meant Territory has developed a robust and adaptable internal rendering capability based on CPU rendering. Peter Eszenyi, creative lead at Territory Studio, goes into more detail: “We try to stick to one solution per project. One of the advantages of our pipeline is that we can accommodate almost any renderer. We use Cinema 4D’s native renderer as well as Arnold and V-ray, as these also work with Maya. Sometimes the project requirements dictate what to use, for example if we need to output deep compositing passes, that narrows the selection of render engines that we can use.”
Looking to the future, Territory Studio doesn’t see any render solution becoming dominant, as that has the potential to cul-de-sac it technically, as Peter explains:
“In my opinion everything is heading towards some sort of standardisation. Shaders will be similar, lighting methods will be similar. Creativity will be the factor that matters, and for us that is the most important thing.”
“There have been quite a few people asking us for Opencl support but, in 99% of the cases, they are AMD users. So in our minds, this is not a CUDA vs Opencl question but an NVIDIA vs AMD one. AMD has been making some aggressive moves in the GPU computer arena recently (and going beyond Opencl) and we’ve been closely following them.”
While one of the key advantages of GPU rendering is the speed and cost savings compared to CPU, many larger studios have been more cautious.
Because many cloud render farms only support Cpu-based render solutions, there can be potential delivery issues if clients need last-minute changes with a Gpu-based solution, as it is harder to scale quickly. That’s not to say that GPU rendering isn’t making the transition to the cloud. OTOY’S Octane utilises its own cloud solution, which is in the early stages of being rolled out. Independent render farms such as Yellowdog are also making GPU services available, as Gareth from Yellowdog reveals: “We were one of the first companies to support cloud GPU rendering with Redshift. We are finding that the usage of Redshift is increasing significantly, with many studios beginning to move from experimentation to using it in their production pipeline.”
crossing the streams
So has the CPU render solution become redundant? Far from it. CPUS and GPUS use vastly different methodologies to get things done, and each has their place. In fact the ideal render solution could use the sequential power of the CPU for complex tasks, combined with the massive parallel power of the GPU to handle multiple instances of simpler tasks.
Indigo Render in V3 offered a ‘Hybrid’ render solution before deciding to go fully GPU for V4, due to the speed increase that could be achieved. Hybrid render solutions may not be as quick as a pure GPU solution, as they are designed to work across both the CPU and GPU. This does give Hybrid engines an advantage because they can potentially work on any machine, from an ultrabook to a multi-gpu and CPU workstation.
One of the most influential Hybrid engines is the Open Source Cycles Render, which has become integrated into Blender. Cycles Render offers an Interactive Render Preview, and the option to use all the GPUS and CPUS available. Cycles Render is also available for Poser and Rhino.
Cinema 4D users have a branch of Cycles Render available through Insydium, the makers of the Cinema 4D plug-in, X-particles. Mike Batchelor from Insydium explains why he wanted to expand from particles to a render engine: “Increasing our product portfolio, with quality software has always
“WE ARE FINDING THE USAGE OF REDSHIFT IS INCREASING SIGNIFICANTLY, WITH MANY STUDIOS USING IT IN THEIR PRODUCTION PIPELINE” Gareth Williams, Yellowdog
been our long-term plan. The Blender Foundation’s Cycles is remarkable, with a supportive global online community.
“Sometimes being from an Open Source community may have the stigma of not being production worthy. However, Cycles has a rich production history. With Cycles 4D you have the deepest support for X-particles, so users can do more advanced rendering effects.
“We brought in a developer who had been working on the early implementation of the bridge. With lots of hard work from the whole team, we launched Cycles 4D in November 2016.”
With the ability to utilise all of the resources of a computer, thus removing the memory allocation issues that can often plague GPU rendering, Hybrid rendering is getting attention from some of the biggest companies in the industry.
Chaos Group has just released V-ray Hybrid as part of V-ray 3.6 for 3ds Max. V-ray Hybrid promises its users greater speed than just using a GPU alone, with the fallback of full CPU capability if the scene gets too big. With V-ray Hybrid, there is the potential to combine the advantages of the traditional Cpu-based V-ray, which is a mainstay of the
industry, with the power of the GPU.
AMD has released the Hybrid Radeon Prorender, which like Cycles Render, is available for free for Blender, 3ds Max, Maya, Rhino and Solidworks, and at the time of writing, will be integrated into Cinema 4D R19 later in 2017. What’s more, it’s written for Opencl so it is hardware agnostic.
Rob Jamieson, Industry Alliance Manager/workstation Graphics for AMD, who is one of the originators of AMD Radeon Prorender, explains more: “Radeon Prorender started as a project to look at what a GPU could do in the next five years with real-time rendering. When we started to delve, we found that the algorithms were really fast on modern GPUS and when we did some tests against CPU versions, it was clear the code that we came up with had more uses than just realtime rendering.
“This code created Radeon Rays, which is a ‘ray caster’ that works on the GPU and is a standalone Api/code. It is being used by other render engines and in other areas. Radeon Rays is the basis of Radeon Prorender.”
3D artist Adrian Wise has switched to AMD Radeon Prorender primarily due to changes in 3ds Max: “Twelve months ago my rendering pipeline was pretty fixed. My small one-
“WHEN WE DID SOME TESTS AGAINST CPU VERSIONS, IT WAS CLEAR THE CODE… HAD MORE USES THAN JUST REAL-TIME RENDERING” Rob Jamieson, AMD
“THE BIGGEST HEADACHE IN THIS INDUSTRY IS COMPATIBILITY” Adrian Wise
man studio produced images for stock libraries. Due to the large format, high-resolution, highquality requirements, and the large volume of images that need to be produced, I ran a small Mental Ray farm locally from my office.
“Since the removal of Mental Ray in 3ds Max as a default component and its subsequent replacement by Arnold, network rendering on my farm now using Arnold comes with an added software cost.
“In response to this I have spent the past six months looking for alternatives. I’ve been using AMD Radeon Prorender for nearly a year now and have just switched my production workflow over to it and returned to a scalable no-cost solution, which also has the added benefit of increasing the quality of my renders at higher speeds.” creating a standard Adrian, like Peter from Territory Studio, would very much like to see a more fundamental change to how artists work, which could then help simplify everyone’s choice of render engine by standardising materials systems.
“The biggest headache in this industry is compatibility.” Adrian says. “It needs standardisation across platforms to make not just assets, but pipelines portable.”
It will be interesting to see how rendering evolves, as GPU and Hybrid options create faster local solutions, and cloud rendering capabilities make pure CPU render engines nearly instantaneous. Couple this with improvements in real-time game engines, and artists now have the power to develop their designs quicker than ever.