|
Written by Michael Lawson, CTO at ArtVPS
|
|
Wednesday, 21 October 2009 |
| Harnessing GPU power might speed up the rendering processes but the real challenge lies in freeing the artist from the render-change-repeat workflow and giving them more time to be creative
Shaderlight delivers significant productivity and creativity improvements over traditional ray-trace rendering engines. It renders intelligent pixels that understand where they fit in a 3D image - and what to do if something changes
At SIGGRAPH this year the talk was all about the GPU and harnessing its power to accelerate ray-tracing. No-one can get enough of it and it seems real-time rendering it has arrived – or has it?
There were several announcements from GPU manufacturers and rendering specialists heralding in this new era in accelerated ray-tracing.
Nvidia launched their ray-tracing engine Optix – a programmable pipeline that offers increased ray-tracing speeds to third-party software developers utilising Nvidia Quadro FX graphics cards.
Harnessing the power of the GPU, Nvidia-owned mental images upped its game with iRay – a progressive renderer based on the CUDA GPU computing architecture. As the incumbent renderer for Autodesk, as well as SolidWorks, Dassault Systèmes and SensAble, it doesn’t look like we’ll have long to wait until GPU-accelerated mental ray appears in our 3D software.
Chaos Software also unveiled a GPU-accelerated version of V-Ray RT. The render engine is still under development and only a part of the software was shown. The demo was also built on the CUDA architecture, but Chaos was clear in its intent to port the application to platform-independent OpenCL. Importantly, both iRay and V-Ray RT are limited to physically accurate renders, which may prove a limitation in those situations where mood or effect rather than realism is the goal.
Caustic Graphics has taken a different approach. They accelerate ray tracing with their CausticOne dedicated rendering hardware and the associated CausticGL API. They have demonstrated significant speed improvements in ray-tracing a 3ds Max scene using Brazil and the Caustic One hardware. Although Caustic bought Splutterfish Inc (developers of Brazil), they assert that they are not in the business of developing rendering software – their aim is for third-party developers to integrate their technology into existing renderers.
Working smarter
These are all natural developments. Ray-trace rendering is processer-hungry and it makes sense to utilise whatever resources are available. At ArtVPS we know all about hardware acceleration, but is faster raytracing really enough in today’s production environment?
Speed isn’t everything. It may mean you can render the entire image quicker but what if you don’t want to start from scratch after one simple change. GPU acceleration is a great step forward but it is not going to change workflow, and that is what’s needed if 3D artists are going to be able to become more productive.
Until now a typical workflow for 3D artists and designers would involve spending time creating preview renders before sending the image off to render. Any last-minute changes would mean starting the entire process again, an iterative and time-consuming approach, which eats into project timelines and inhibits creativity.
From a workflow perspective the user is always working with a fully rendered image
Shaderlight delivers significant productivity and creativity improvements over traditional ray-trace rendering engines. It renders intelligent pixels that understand where they fit in a 3D image - and what to do if something changes. This enables users to change key attributes of a full resolution image at any stage of the rendering process without ever having to re-render. When changes are made to materials, environments, lights or textures, the information embedded in each pixel is used to update the image without the need to re-render. We call these MELT changes and this approach dramatically improves efficiency and project turnaround. Users are not only able see the progressive refinement of a scene when objects or camera angles are altered but can also make interactive changes to the MELT elements on full quality 3D images. Additionally, a new approach to global illumination (GI) enables light, colour and intensity to be updated on screen without the need to re-render. From a workflow perspective the user is always working with a fully rendered image.
We developed Shaderlight after extensive research, which showed the top pain issues for 3D artists to be rendering speed, lighting and material set-up and fine-tuning an image. Shaderlight overcomes these problems and meets the demand for an interactive solution that frees the user from the frustrating render-change-repeat workflow, giving them more time to be creative
We are well aware of the added value of GPU acceleration and while this current release will run on the CPU we are not ignoring the GPU. Our architecture is designed to be flexible and future versions of Shaderlight will support hardware-based acceleration.
So while most rendering specialists at SIGGRAPH were talking about speed, our own discussions about interactive rendering opened a few eyes to the idea of working smarter rather than harder. Encouraged and enthused, we are keen to continue working closely with our technology partners to ensure that as many creative 3D professionals as possible can take advantage of Shaderlight.
If you find regular news, features and product announcements useful, why not sign up to the 'Mechanical Design' weekly newsletter. Simply click HERE to register.
|
