September 7, 2019in News 41 Likes

Deep Learning Training Efficiency

Deep Learning Training Efficiency

Deep Learning Requires Learning

From technology to manufacturing, from healthcare to insurance, discussions surrounding the impact of artificial intelligence on a host of industries continue to gather momentum. At present, a significant amount of attention is focused on training accurate models (based on specific data sets) as quickly as possible. And because AI involves a complex matrix of operations and the vast number of iterations, the performance of the computing platform is very critical—and we’re not just talking about GPUs. They’re important, of course, but there are also a multitude of other critical factors that need to be addressed, beginning with GPU processing power vs. data storage performance:

  • Understanding the characteristics of your chosen DL framework

  • Dual CPU systems present special challenges, so using two CPUs rather than one is only half the story

  • Local storage is a major source of bottlenecks in deep learning applications

  • Selecting the right NVMe drives

  • With data storage and multi-GPU systems, adding GPUs is an effective way of accelerating DNN model development

  • A clear understanding of the inherent latency of your platform

And because a mismatched storage sub-system can leave the GPU (or multiple GPUs) idle, leading to a severe loss of training efficiency, data richness and DNN training topics like aligning data type and storage sub-system performance or how to solve a latency problem, are also essential.

In order to be successful, data scientists working to achieve the fastest possible time-to-model need to fully understand the precise hardware configuration necessary for their specific data type. That’s why the new BOXX whitepaper, Deep Learning Training Efficiency: It’s Not Just About GPUs, is essential reading. From GPU processing and data storage performance to data richness and DNN training performance, the deep learning infrastructure experts at BOXX and Cirrascale Cloud Services provide the critical information you need to configure the right AI computing platform.

September 7, 2019in News 41 Likes

Animal motion capture data developments & the impact on animation

Animal motion capture data developments & the impact on animation

Animal Motion Capture Data Supports Realistic Animal Avatar Animation

How do studios accurately capture animal movement? Animators and moviemakers may have a new avenue with fresh motion capture research from the University of Bath’s Centre for Analysis of Motion, Entertainment Research & Applications (CAMERA), which has been collecting motion data from different dog breeds. This could be the next step towards improved and more precise visual effects for animals in movies, video games, and other media.

The entertainment industry has been trying to correctly animate and capture animal movements for close to a century. Walt Disney’s breakthroughs in short and feature-length films like Steamboat Willie, Snow White, and the Seven Dwarfs, and Dumbo paved the way for semi-realistic portrayals of animals. State-of-the-art animation techniques helped to anthropomorphize animals in film.

In early Disney films, producers used a method called rotoscoping to shoot actors playing out parts and then tracing over that motion during animation. Rotoscoping is the forerunner to motion capture and is still in use today. So how did we get to where we are today with motion capture?

The History of Motion Capture:

The first fully digital protagonist in a motion picture was Jar Jar Binks from Star Wars Episode I: The Phantom Menace. Motion capture was also used in The Mummy and The Gladiator for specific scenes, but the technology was cost-prohibitive at the beginning. It quickly became indispensable for certain films, but animation studios were divided on the value of its benefits. 2006 Oscar winner Happy Feet used the technology extensively, while 2007’s Ratatouille proudly claimed it was “100% Pure Animation—No Motion Capture.”

Regardless of differences in opinion, the technology and research behind motion capture developed further. Modern motion capture was used for roles like Caesar (an ape) in Dawn of the Planet of the Apes and for Smaug (a dragon) in The Hobbit: The Desolation of Smaug. These animal or animal-like roles have used motion capture primarily provided by humans. The next frontier is motion capture which accurately captures how animals move. Instead of using rotoscoping, observing animals, or anthropomorphizing them, we can display how they move in reality with the help of the research in progress at the University of Bath.

Motion Capture Research from the University of Bath

Animated animals have been present in film for close to a century, but it’s been up to the animators’ senses of motion and the application of rotoscoping to determine movement. The advent of a new technique from CAMERA uses motion capture from different breeds of dogs to create realistic movements. They used animals from the Bath Cats and Dogs Home to create a library of real animal data. The goal is to be able to translate human motion capture through this library to create more accurate dog movement.

The dogs wear coats with reflective markers, and their movement is recorded by a specialized camera to record their 3D position. Animation studios can use this new technology for accurate animal animation. Using the latest animation workstations from BOXX, movie and gaming studios can utilize this new programming to create more realistic animation than ever before. Aggregate motion data can create surprisingly accurate guesses of how animals move, which gives 3D motion software a solid idea of the limits and ranges of animal movement.

Humans can then imitate animal movements using existing motion capture technology and feed that information back to the motion graphics software. The software approximates human movements with its library of animal movements, then builds an animation based on a premade 3D model.

This research project can potentially change the way motion capture works for animal roles in film, gaming, and other visual industries by continually evolving a computational understanding of animal movement and taking full control of that movement using human actors. It’s the evolution of animal motion capture.

September 7, 2019in News 41 Likes

Learn about the latest updates & improvements in KeyShot 7.2

Learn about the latest updates & improvements in KeyShot 7.2

KeyShot 7.2: The Biggest User Experience Improvements

KeyShot, the 3D rendering and animation application from Luxion, allows product and graphic designers to quickly transform 3D models into photorealistic images. Now on its seventh iteration, KeyShot has streamlined its interface and introduced several exciting new features to improve the overall user experience. Here are a few welcome changes that the latest update, KeyShot 7.2, brings to its users:

Better Appearance on the Screen

The new KeyShot 7.2 offers HiDPI support, which means that icons, text, and the real-time rendering view look crisper. The HiDPI support also includes the ability to respect the physical display size you’re using. If you’re using a 4K display with system scaling applied, KeyShot 7.2 will take the pixel size into account, and you’ll still see a high-quality image on your screen.

Improved Scene Tree Reordering

While the makers of KeyShot still recommend arranging your models or parts in CAD first, they’ve made it less cumbersome to reorder elements in the Scene Tree. KeyShot 7.2 has made improvements to the re-ordering indicators feature so that it’s easier to drag a model or part to a different level of the Scene Tree. It’s now also possible to reorder Model Sets in the Scene Tree sidebar or KeyShot Configurator.

Improved Rendering

KeyShot 7.2 lets users load complex scenes significantly faster than before. Luxion has also made improvements to KeyShot’s Rounded Edges feature, which will be particularly useful for designers who need to simulate realistic fillets on parts. Users will notice improvements to Ground Shadows with Motion Blur as well.

KeyShot Configurator Changes

KeyShot Configurator is a relatively new feature available with KeyShot Pro. It can be used to present model variations for design reviews or interactive point-of-sale displays (if a furniture retailer wanted to let shoppers view the same chair in different colors and materials, for example). The KeyShot 7.2 update has made it easy to render out all model variations in the Configurator with one click. Users can also set labels to show as a separate color in a Clown Pass, which makes selection and masking in Photoshop easier.

Bug Fixes

KeyShot 7.2 has resolved several bugs from previous releases. Notable bug fixes include the resolution of crashes that could occur when duplicating Camera, updating geometry in multiple Model Sets, and rendering with Motion Blur. KeyShot 7.2 also resolved an issue with pixel filtering in the real-time view that could decrease the sharpness of an image.

Powerful Workstations for Advanced Rendering

KeyShot relies on a multi-threaded CPU. That means that the more cores and threads you have in your workstation, the faster KeyShot runs. You can start improving your KeyShot 7.2 performance with a safely overclocked workstation and renderPRO from BOXX. Our workstation solutions help you avoid bottlenecks with rendering so that you can get more done in less time.

 

So which BOXX workstations would we recommend for Keyshot?

The APEXX D5 is our most advanced workstation platform. With support for up to five GPUs and powered by dual Intel Xeon SP Processors with up to 56 cores, the APEXX D5 can support a variety of professional workflows.

The APEXX W5 is our most advanced workstation platform ideal for rendering, simulation Deep Learning and other professional GPU accelerated compute applications. The system is optimized for GPU centric workflows.
Whether you’re editing feature films or deploying cutting-edge VR experiences, the APEXX X3, with support for up to two* graphics cards and 128GB of memory, is an ideal platform for extreme multi-tasking in your complex production pipeline.

 

September 2, 2019in News 41 Likes

Unity & Autodesk Collaborate to Make 3D Easier

Unity & Autodesk Collaborate to Make 3D Easier

 

Unity Technologies is well known for creating the extremely flexible 2D, 3D, and VR experience creation tool Unity, and they have just teamed up with a fellow leader in 3D design and engineering software, Autodesk, to help streamline the workflows of companies who work with their products.

It’s not unusual for creators to jump from software application to application to perform certain tasks more robustly or easily. Where some software may be designed for strength in certain aspects of 2D and 3D design, others may need to make sacrifices in the same area to be strong somewhere else.

In answer to the modern workflow problem of application jumping, Unity and Autodesk have revealed a first look at improved application interaction between Unity and 3ds Max/Maya at the Unite Austin 2017 Conference.

Collaboration Launches with Unity 2017.2

Unity 2017 promises to offer an ever-evolving creation engine for gaming and entertainment, giving artists and designers a shot at improved workflow. It’s already one of the most widely used gaming engines, accounting for more than 50% of the base of new mobile games and an astounding 66% of today’s VR experiences.

Collaborating with Autodesk will give Unity users access to custom Autodesk properties, such as the interactive physically-based shader. Customers can likely expect more seamless integration as updates continue to roll out, including new and improved FBX import and export capabilities.

Improved import and export for FBX files could help increase flexibility for those working between applications or teams who work side-by-side with different software. It’s only the beginning – further collaboration could spark technology partnerships that bring incredible new capabilities to 2D and 3D design tools.

Though the results of the collaboration won’t apply retroactively to older products, continued improvements will be available for users of Unity 2017.2 and beyond. The amount of asset sharing and integration that is happening already is extremely promising, and the 2D and 3D design industry can likely expect workflows to continue to enable the realization of bigger and better ideas.

For more information about BOXX’s workstations optimized for 3ds Max, review our solutions page.

September 2, 2019in News 41 Likes

Develop3D Workstation Report – BOXX APEXX S3

Develop3D Workstation Report – BOXX APEXX S3

 

“With an overclocked six core ‘Coffee Lake’ Core i7 CPU and a well engineered compact chassis, this is an impressive CAD workstation but it does come at a premium,” writes Greg Corke.

When we started DEVELOP3D in 2008, almost all CAD workstations came with a quad-core CPU. If you wanted more cores, say to cut render times, you needed a specialist CPU and deeper pockets. You also had to accept that performance in single-threaded applications like CAD would fall, due to a drop in GHz.

Ten years later and things have finally changed. The new Intel ‘Coffee Lake’ Core i7-8700K is the first mainstream Intel processor to feature six cores. And despite having 50% more than its predecessor, the ‘Kaby Lake’ Core i7-7700K, it should also deliver a small performance improvement in CAD.

The Core i7-8700K has a base frequency of 3. 7G Hz, at which all six cores can run, but is designed to Turbo up to 4.7GHz in lightly threaded workflows. This is a much wider frequency range than the ‘Kaby Lake quad core i7-7700K (4.2GHz- 4.5GHz) but becomes irrelevant in the APEXX S3, the latest workstation from custom system builder BOXX.

BOXX has overclocked the Core i7-8700K so all six HyperThreaded cores run at 4.8GHz, all of the time. This makes a big difference in CPU rendering applications.

The APEXX S3 completed our Luxion KeyShot render test 38% faster than the overclocked 4.6GHz Core i7-7700K Scan 3XS workstation we reviewed in March (tinyurl.com/3XS-D3D). Previously, in order to get this level of rendering performance, you had to go for a specialist overclocked Intel CPU with 8 cores or more, which was significantly more expensive and has a lower clock speed.

This is a welcome boost to CAD users who rely on rendering as a key part of their workflow. The emphasis on CAD is important here, as this where the APEXX S3 excels. Indeed, it set a new record in our SolidWorks IGES export test, which uses a single CPU core to process the parts in a complex CAD assembly.

Our test machine was kitted out with 32GB DDR4-2666MHz memory, which should be a match for very complex CAD models. But if you frequently use multiple applications as part of your product development workflow, then it can go up to 64GB, distributed across 8 DIMMs.

In order to push the Core i7-8700K beyond its stock clock speed, BOXX has used an enterprise-class closed loop water cooling solution which connects via a flexible pipe to a large radiator at the front of the machine, cooled by two fans. With air drawn in at the front and expelled at the rear of the machine, this does mean some of the heat extracted from the CPU is pushed back into the chassis. However, while fan noise was noticeable, we found it to be perfectly acceptable throughout all of our testing – although the RPM of the fans cycled up and down every 10 seconds or so during our KeyShot renders which some might find annoying (this didn’t happen in V-Ray or SolidWorks Photo View 360 where it was consistent).

The chassis is nicely designed with an angled I/0 panel front and top giving easy access to four USB 3.1 ports (there are also four USB 3.1 and two USB 3.0 at the rear). This leaves the entire front fade free for uninterrupted airflow. To stop dust from getting in there is a removable air filter.

The chassis is custom-made in Texas from aircraft-grade aluminum. I’m no aerospace engineer, but it’s obvious that it has a strength and rigidity beyond that of off-the-shelf cases. It’s also notably smaller than many of the expandable overclocked machines we review at DEVELOP3D.

One of the reasons for this is that BOXX has put two 3.5-inch drive bays at the top of the chassis instead of the front. There is also no optical drive. Our test machine came with a single 512GB Samsung SSD 960 Pro, one of the best workstation-class M.2 NVMe SSDs on the market. It is housed directly on the AS Rock Z370 Taichi motherboard, which also has two additional M.2 slots, so there are plenty of storage options.

Graphics in our test machine was provided by a single Nvidia Quadro P2000 GPU, an excellent choice for 3D CAD. The APEXX S3 managed to smoothly navigate every SolidWorks and PTC Creo model we threw at it at FHD resolution, even with real-time lighting, shadows and reflections enabled.

We started to see the limitations of the CAD-focused GPU in-game engine design viz application LumenRT where frame rates dropped to 13.5 FPS with one dataset. This will only get slower at 4K resolutions, so those who take their real-time visualization more seriously, or want to take their designs into VR, should consider a Quadro P4000 (which we also tested) or even a Quadro PS000, Quadro P6000 or AMD Radeon Pro WX 9100.

The APEXX S3 can actually accommodate up to three GPUs (two double slots, one single slot) thanks to its 650W PSU so there are plenty of options for those using multi-GPU aware applications like Autodesk VRED or GPU renderers like V-Ray RT, Solid Works Visualize or Radeon Pro Render.

CONCLUSION

Overall, the APEXX S3 is a very impressive high-performance CAD workstation that is built around a solid, well-engineered chassis. There is a downside though. Quality like this comes at a premium and our review machine will set you back £2,899 (with a Quadro P2000) and £3,299 (with a Quadro P4000). This is notably more than a similarly specced machine from other custom workstation manufacturers in the UK. BOXX justifies this price tag not only by the quality of its machines but also by the quality of its tech support, which in the past we have found to be very knowledgeable in a number of CAD and rendering applications. There are many BOXX customers that feel this is a price worth paying.

 
September 2, 2019in News 41 Likes

Hybrid Rendering for V-Ray 3.6 + Maya Opens New Doors

Chaos Group has long had one of the leading rendering software pieces in the market, but it is poised in 2018 to take advantage of the ability of GPUs and CPUs to work in tandem to maximize one another’s potential.

V-Ray for Maya, which has been a part of major entertainment motion picture projects such as Stranger Things, Spider Man: Homecoming, and Guardians of the Galaxy 2, is receiving a hybrid rendering upgrade. For high-end video effects production, that means that GPUs and CPUs will not (to as significant of a degree) limit the performance of one another.

Powerful rendering technology already exists but getting software to work together with hardware has been a tough obstacle. NVIDIA’s CUDA-powered GPU rendering technology works simultaneously with CPUs to help streamline multi-threaded rendering tasks, but only if the software is willing to do the same.

V-Ray for Maya takes full advantage of NVIDIA’s hardware setup and gives designers and artists the flexibility to choose what technology renders their scenes. But that’s not all – the V-Ray 3.6 update also comes with a few extra performance boosts:

  • NVIDA NVLink – Allows for shared GPU memory on graphics cards compatible with NVLink technology, which has the potential to eliminate recurring memory errors while rendering.

  • MDL Support – Gives V-Ray users access to the NVIDIA universal material format, broadening the range of material-creation software usable in Maya.

  • Viewport 2.0 Changes – V-Ray now supports lights in Viewport 2.0, which helps streamline light setup and improves support for environmental reflections and V-Ray-specific materials.

  • Cryptomatte Addition – Generates ID mattes automatically to help speed workflow; allows quick access to sweeping changes in such effects as transparency, depth of field, and motion blur.

  • Full Light Selected Render Element – Improves accuracy of light mixing by allowing for single or multiple lights to be selected and rendered individually.

These changes have been made to answer to rapidly evolving workflows in the entertainment industry. V-Ray 3.6 is another major step toward total rendering efficiency for major projects and is helping push the industry toward a universal standard. Check out our advice on GPU rendering with BOXX