After reviewing Octane Render we decided to make an interview with OTOY to get known the company which stays behind the development of the renderer and as well ask some questions about the renderer itself. OTOY took over the development of Octane Render from a New Zealand company called Refractive Software and continued the development on it, bringing new features and enhancement to the renderer. Moreover, Octane Render is the world's first GPU based, un-biased, physically based renderer. And from my opinion is worth to try. We were asking questions Sam Lapere, product manager for Octane Render at OTOY New Zealand and we would like to thank him one more time for taking the time for answering our questions.
Check also the video at the bottom in which are presented works did with Octane Render by Erik Jansson.
Q: Can you tell us something about the history of OTOY?
OTOY was founded by Jules Urbach, Alissa Grainger and Malcolm Taylor in 2009 and has two main focuses: delivering remote services on the cloud like server side rendering, streaming games and 3D content creation apps and the LightStage, which is an extremely high quality capturing environment able to make 100% photorealistic representations of actors.
Q: OTOY is focused mainly on cloud Technologies. What was the reason to take over the development of Octane Render?
OTOY has experience with building clusters of GPUs and server side rendering for games. Octane Render had piqued our interest early in its beta development phase because of its speed, quality and ease of use, but it required a high end GPU to fully appreciate its power. We want to democratize Octane Render so everyone can enjoy very fast photorealistic 3D rendering on any device, even iPads, without having to worry about the GPU or CPU they have inside.
Q: Octane Render is the first unbiased renderer purely based on the GPU. Why did you choose this approach? Why not a hybrid CPU-GPU?
Using the CPU in hybrid rendering mode does not provide much benefit unless you have a very powerful one. Hybrid rendering means that CPU rendering must be synchronized with GPU rendering which is almost always less efficient than pure GPU rendering, because the GPU has to wait for the CPU and starts idling. Our goal is to have the fastest renderer on the market, so we focus all our efforts and developer resources on the GPU. It's also much easier and more effective use of resources to concentrate on one platform only.
Q: What are the advantages of Octane Render comparing to other renderers on the market?
Octane Render's main advantages are its speed, rendering quality and ease of use: Octane is the fastest renderer on the market, it's also easy and fast to set up your scene, camera, lights and materials interactively and thanks to the progressive rendering you get an immediate preview of the final quality image.
Q: Comparing the price of Octane Render among other renderers the price is very low. What were the reasons to set the price so low.
Good point. Even though Octane competes with renderers that cost ten times as much, we decided to keep the price very low because we want Octane Render to be enjoyed by as many people as possible. We target not only 3D professionals like architects and product designers, but also freelancers and hobbyists.
Q: Where can be Octane Render mostly used? (Visualizations, Architecture, VFX …)
Octane can be used for everything 3D related: special effects for films, 3D animation, TV commercials, 3D stills, architecture, product design and automotive visualization.
Q: If we compare some Intel i7 CPU which has 4 cores and 8 threads to a Nvidia GPU which can have hundreds of CUDA cores does this means that the performance is rising linearly with the number of cores? Is there any correlation between the CPU cores and GPU cores?
There is no correlation between CPU and GPU cores. GPU cores are designed to be smaller and simpler then Cpu cores so you can fit more of them on one board. Also, GPU core counts cannot easily be compared accros GPU generations, for example a high end Kepler GPU like the GTX 680 has about 3x as much cores as the GTX 580 which is based on Fermi, but each of the Kepler cores is a lot less powerful than a Fermi core, which makes Kepler about as fast as Fermi. Nevertheless, Octane Render scales extremely well with the number of GPUs and the number of cores per GPU, for example, when using 8 GPUs, we observe a speedup of 7.92x.
Q: Looking a bit into the past and comparing the increase of performance by CPUs and GPUs we could say that the performance by GPUs is increasing faster than by CPUs. Was this one of the reasons why to develop a GPU based renderer?
Absolutely. We have closely followed the research being done with GPU ray tracing for a while. Around Siggraph 2009, it became clear that the GPU would become the best platform to develop a renderer on. The massive number of cores on today's GPU combined with the “embarrassingly parallel” and fine-grained nature of the ray tracing algorithms used by Octane Render result in a 10 to 20x speedup when using the GPU compared to the CPU, and we think we can make the gap even bigger soon.
Q: What is the main factor by the GPU and rendering with Octane Render? Can the memory of the GPU increase the performance or the main factor are the CUDA cores?
The amount of memory on the graphics card has no impact on the rendering performance, it just allows to render larger scenes with more and larger textures. The specific GPU architecture, number of cores and GPUs and core clock are the main performance determining factors.
Q: The current GPUs have different architecture starting from Tesla, Kepler or Fermi. Are all compatible with Octane Render? Could you tell our user some difference between them?
Yes, Octane Render is compatible with all the CUDA capable architectures, starting from the Tesla architecture (e.g. GTX 8xxx, GTX 9xxx, GTX 2xx) to Fermi (e.g. GTX 4xx and 5xx) and Kepler (GTX 6xx). Octane Render is also compatible with GeForce, Quadro and Tesla cards. The Fermi cards are incredibly powerful compared to the previous generation, but also consumed a lot of power. The Kepler cards generally have more memory, consume less power and allow for much more textures to be used with Octane Render than Fermi.
Q: What could be the ideal workstation for Octane Render? It’s better to use the Quadro line or Geforce line from Nvidia?
We currently recommend our users to have a dedicated low end GPU for the GUI and OS (this can even be Intel integrated graphics) and one or more GeForce GTX 680 or GTX 690 GPUs for rendering. The GTX 680 has up to 4 GB RAM, it runs efficiently and silently and Octane renders very fast on this card. The GTX 690 is twice as fast as the GTX 680 in Octane and using multiple GPUs (if your budget allows it) will significantly enhance the rendering experience. We haven't seen any benefit from using Quadro compared to GeForce besides having more video memory.
Q: Octane Render can access a user build material library, do you also contribute to the library or its only user based?
Yes, we have artists in our team that upload materials (for example, we recently made a very complex and realistic skin shader that uses subsurface scattering) to the online material library so other users can apply it immediately to their models.
Q: Are you proud on any particularly moment when Octane Render was used?
It may sound a bit cheesy, but we are often proud when we see an extraordinary render or animation. If I would have to pick one that I am particularly impressed with, I would say an architectural visualization made by WakYak