As SIGGRAPH 2018 has wrapped up in Vancouver, Nodal brings you the latest product release announcements from the event that will impact the film, TV, and advertising industry.
OctaneRender 2018.1 Features Announced
Developer Otoy showed off some of its latest features for its upcoming OctaneRender 2018.1 release, including its Vectron and Spectron technology. OctaneRender 2018.1 (effectively version 5.0 but with a new naming scheme), will feature previously-announced trace sets, as well as an improved system for managing mattes and AOVs.
The developer also plans on vertex attribute support to better integrate with tools such as Houdini, Cinema 4D R20, TurbulenceFD, and X-Particles, to name a few.
Standing for ‘vector-polygon’, Vectron is being referred to as a new procedural ‘uberprimitive’ meant to provide a more efficient representation of scene geometry than current volumes and meshes - to the point where Otoy claims Vectrons will not use any VRAM and has a ‘zero memory footprint’.
The technology is still brand-new so how Vectrons will integrate into modern production environments is an open question. However, they do seem to work well with abstract and fractal effects and may end up being a more efficient in-between format for rendering complex scene geometry.
Spectron is the sister technology to Vectron and is a procedural volumetric lighting system. Otoy claims that the technology will provide noise-free and near real-time rendering while mimicking features of physical lights such as barn doors, gels, and blockers. Other features include support for light polarization, diffraction, complex layered materials, and more.
Other promised features for OctaneRender 2018.1 include a new cel shader, plus OpenVDP support for volumes.
The ‘ubermaterial’ released in OctaneRender 4.0 will gain subsurface scattering support.
New shading options and spectral denoiser for hair, as well as AI-based viewport image upsampling.
Performance updates include enhancements to the AI Scene system and a feature that allows meshes to have unlimited primitive counts.
For a more in-depth look at the feature list for Otoy’s upcoming release, check out CGChannel’s write-up.
Bear in mind that the release windows for promised features may slip and previously announced tools in past releases have not always arrived as planned. Still, this major release promises to add some exciting new tools to the GPU rendering toolkit.
NVIDIA Unveils Turing GPU Architecture
Graphics card manufacturer NVIDIA announced their highly-anticipated next generation of GPUs built around their new Turing architecture. While the full capabilities of the new cards have yet to be seen, the new features look to boost the cards’ rendering capabilities significantly.
With Turing, NVIDIA is looking to utilize the strengths of both ray tracing and traditional rasterization to deliver fast, efficient hybrid rendering. The Turing cards will feature a dedicated ‘RT core’, which will serve as a dedicated ray tracing processor. NVIDIA is claiming these cards will be able to cast up to 10 billion rays per second, which at first blush would offer an up to a 25-fold improvement over older Pascal cards without ray trace acceleration.
The Turing cards will also carry over the ‘tensor cores’ from NVIDIA’s Volta model, which not only help accelerate ray tracing, but seek to reduce the number of rays required by using AI-driven denoising to clean up a render.
Overall the company is suggesting that these new cards will offer up to 6 times the performance of its Pascal GPUs, though it remains unclear whether this factor will hold up in all render situations. Even with dedicated RT cores, ray tracing is still exceptionally resource-intensive.
As part of the announcement, NVIDIA confirmed that its Turing cards will support the newest generation of graphics memory, GDDR6. GDDR6 supports up to 16Gbps per pin of memory bandwidth, making it twice as fast as NVIDIA’s GDDR5 cards (and still a 40% improvement over the GDDR5X onboard memory). GDDR6 continues to improve upon features implemented in the previous GDDR5X model, including a lower operating voltage (1.35v) and offering two memory channels per chip, allowing for a significant support of parallel operations.
Coming Late 2018
NVIDIA announced that the first three cards built around the Turing GPU, the Quadro RTX 8000, RTX 6000, and RTX 5000, will be available in late 2018. Though NVIDIA’s next generation of consumer cards have not yet been announced, it’s likely that they will arrive even earlier, as they have lesser validation requirements. Cost for the new cards will be high upon release, as the RTX 8000 is set to retail at $10,000.
Check out AnandTech’s deep dive on all the new features and performance numbers for the new Turing cards.
AMD’s Second-Gen Ryzen Threadripper CPU Released
AMD announced the release dates and model numbers of its next generation of ‘Threadripper’ processors, with the 32-Core Threadripper 2990WX already available. The 16-Core 2950WX is set to release August 31, 2018, with 24- and 12-Core models arriving in October 2018.
The new generation will be backward compatible with the older X399 platform and the Socket TR4 used with the first generation and they will feature support for quad-channel DDR4 memory.
The 2990WX is set to run with a 3.0GHz clock speed, with a 4.2GHz ‘boost’ speed. It also sports 64MB of L3 cache. The 32-Core 2990WX model is set to retail at $1,799, while the 2950WX is set for an $899 price point.
For a more breakdown of all the upcoming Threadripper CPU specs and pricing, check out Neowin.net’s coverage of the announcement. For a technical deep dive, Anandtech has a complete review of the 32- and 16-Core CPUs.
AMD Announces Radeon Pro WX 8200 for Less Than $1,000
The Radeon Pro line of GPUs is AMD’s competitor to the popular Quadro series by NVIDIA, and AMD has just announced a new offering for content creators on a budget. The WX 8200 model performs favorably compared to the Quadro P4000, specifically in common creative tools such as Maya, Nuke, and Premiere, though it is still outstripped by the more expensive Quadro P5000.
While the WX 8200 represents a performance hit compared to AMD’s WX 9100 released last year, it also sets you back significantly less than the WX 9100’s $2,200 price tag. The WX 8200 boasts 3,584 GPU cores (compared to the WX 9100’s 4,096) and ships with 8GB memory capacity.
Overall the WX 8200 represents a slim, budget-conscious solution as AMD continues to move further into the professional GPU market so long dominated by NVIDIA. The new card will retail for $999 and is set to be available in early September.
For more coverage on AMD’s new offering, check out ExtremeTech’s take on the WX 8200. Appual’s breaks down the technical specifications of the new card. Finally, XtremeGaminerd also has coverage on the product from a consumer standpoint.
Pixar’s RenderMan 22 Released
Pixar Animation Studios celebrated their 30th anniversary by releasing RenderMan 22, the latest installment of their storied rendering software. The new version brings in several features intended to speed up the rendering workflow for professionals; Live Rendering is specifically meant to bring continuous photoreal renders directly into your 3D application of choice.
By introducing always-on live rendering capability to their renderer, Pixar has made it easier for professionals to collaborate all across the pipeline - from modeling and animation to lighting and shading. This feature, made possible via a new Universal Scene Description (USD) based method for handling scenes, gives artists constant feedback and improves iteration speed. RenderMan tools for Maya, Houdini, and Katana have been updated to accommodate this live rendering technology.
Other new features include improvements to lighting technology originally used in Pixar’s films such as The Incredibles 2 and Finding Dory. This technology, called Pixar Unified, allows for higher image quality and faster production times. The tool allows for per-light selection of unidirectional or bidirectional path tracing, giving artists more flexibility to choose the best tool for the scene at hand. Pixar Unified also seeks to deliver photorealism without long delays, allowing artists to see changes made to their lighting as they’re made.
Finally, RenderMan 22 promises significant speed increases, which will definitely have an effect on production times. Most of this speed-up has been due to Pixar streamlining legacy code. This architecture redesign reduced the codebase by 30% and should offer nearly double the speed of previous versions.
While RenderMan 22 does not feature Pixar’s announced XPU technology (which will utilize both the system’s CPU and GPU for final renders), the new version is still a major improvement for your rendering pipeline. RenderMan 22 is now available for download for customers on maintenance contracts.
For more details, Graphic Speak has a breakdown of the new technology and features.
Chaos Group’s New Real-Time Ray Tracing Technology
Leveraging the RT cores in the Turing-based NVIDIA Quadro RTX GPUs described earlier, Chaos Group unveiled Project Lavina, a new technology with the goal of delivering photorealistic real-time ray tracing. Successfully producing real-time ray tracing would be a massive achievement in computer graphics, and Lavina could have significant impact in visual quality for real-time gaming and VR.
Revealed at a tech demo at SIGGRAPH, Lavina was shown to handle complex scenes including a massive 3D forest and several architectural environments at 24-30 frames per second at standard HD resolution. Lavina is allegedly capable of handling scenes with over 300 billion triangles without loss of detail.
Lavina simplifies the process of delivering game engine assets via direct compatibility and translation of V-Ray assets, rather than the rebuilding and optimization usually required. This means that, when a scene is loaded, the user can explore it just as they would in a game, complete with physically accurate lighting, global illumination, and reflections.
For more coverage on Chaos Group’s announcement, check out this article by Animation World Network.
Mo-Sys Displays New StarTrakcerVFX Tool
Along with HP and NVIDIA, Mo-Sys demonstrated its new StarTrackerVFX optical camera tracking system as part of Waskul Entertainment’s StudioXperience broadcast space at SIGGRAPH. This system combines real-world footage and virtual worlds in real-time, reducing the time and budget required for post-production.
In the demonstration, Mo-Sys showed a real-time composite of interviews filmed against a green screen, introducing an environment rendered by the Unreal Engine. The footage and camera tracking data was fed into the Nuke compositor, where VFX could be added in a matter of minutes.
The new tool features a direct plugin for Unreal that enables users to film green screen footage and easily import them into photorealistic virtual environments. Camera tracking data can also be exported via FBX format for use in post-production.
The tracker works by attaching an upward-facing sensor to your physical camera. Then, reflective stickers (‘stars’) are attached to the ceiling and used as reference for any camera movements. The benefits of these stars is that they’re inexpensive, easy to install, and are more resilient against changes in light levels compared to forward-facing sensors. Once the tracker is calibrated against the stars your camera’s position and movements can then be properly tracked.
Broadcast Beat has more information about the initial announcement of Mo-Sys’ SIGGRAPH demonstration.
For more complete coverage of all the news that came out of SIGGRAPH 2018, check out their official site here.
Digital Arts Online also has a breakdown of some of the key announcements.
As always Nodal will keep you up to date on the latest news and announcements from developers across the creative industry. If you have any questions about integrating these new tools at your facility, contact us!