What do VR applications, artificial intelligence for product development, and autonomous survey drones have in common? A need for fast, multiple GPU workstations.

It was a multi-GPU feast for the eyes in the Demo Center during the 2016 Fujitsu Forum that attracted over 12,000 visitors to Munich Germany.  From the business-oriented virtual reality demonstration to the 3D photogrammetry surveying application to the AI-based engineering assistant, the common thread flowed in the direction of powerful multi-GPU workstations.

Fujitsu CELSIUS VR-Ready Workstations

The virtual reality market appears to be growing according to the workstation experts at Fujitsu. The combination of powerful workstations, fast professional graphics and very capable head-mounted displays (HMD) for virtual reality has simultaneously increased the number of use-cases for VR and reduced the expense for a professional VR solution.

The Global Product Marketing Manager for CELSIUS workstations, Natalie Walther, explains that virtual reality in industrial applications is not new, but with software and hardware developments in the last few years, the market is growing.

Ms Walther explains that VR promotes a holistic development process, and one that can be valuable in a range of industries including automotive, aerospace, architecture, manufacturing, healthcare, and consumer products.

In order to have a proper VR system, there are four key factors to pay attention to.

The most obvious is the display system. It must be capable of recreating the 3D environment in a physical 3D space. For many applications, this has required a physical 3D space with the 3D world projected in stereo images on the walls, floor, and sometimes, even overhead. The user or users wear 3D stereo glasses and the system may also include a means of tracking the movement of the user.

A viable and affordable alternative today is the use of a Head-Mounted Display or HMD. In an industrial VR environment, it is essential to have a quality HMD. The display must support a very high refresh rate of 90 Hz or better. It must have a field of view that is 110° or greater.  And then it must be comfortable to wear.

Naturally the software is a critical part of a VR solution. Not only must the software environment support the design and engineering environment with easy integration of data from the CAD and PLM systems, it must also deliver a high-performance virtual reality graphics environment. This typically means integrating design data from other tools into a specialized VR engine.

NVIDIA Quadro M6000
High-performance graphics, like NVIDIA Quadro, are supporting customers who need optimally performing systems in VR, AI, and data analysis.

It comes as no surprise that the graphics system is critical. VR environments need more than 90 images per second for two eyes. In an industrial setting, the design environment is often non-trivial and often, it is undesirable to reduce the fidelity of design data.  This combination of requirements leads to a virtual reality environment in industry that has very high graphics processing demands.

Last-but-not-least, a high-performance VR environment needs a high-performance workstation that brings the HMD, VR applications and graphics performance together into a correctly functioning virtual reality.  Fujitsu and NVIDIA have cooperated to create two “VR-Ready” graphics workstations. One is the CELSIUS M740 workstation that can be configured with two high-performance NVIDIA Quadro graphics cards. Another is the CELSIUS M940 that is a dual-CPU workstation and can be configured with three ultra-high-performance NVIDIA Quadro GPUs like the Quadro P6000 with 24GB of on-board graphics memory.

In summary, a professional VR environment links professional engineering software and design data with a virtual reality graphics engine running on a workstation with professional graphics and a professional HMD.  Taken together, such a VR environment delivers high frame-rates with complex design data.

Artificial Intelligence in Design

What engineer would not want to have a fast, smart and not-too-talkative assistant? This could become reality as Fujitsu is planning new artificial intelligence technology for the Fujitsu Engineering Cloud services in 2017.

Serban Georgescu provided an early look at this design-support technology coming from the Fujitsu Laboratories. One example integrates design expertise into a machine- learning environment. The solution on display determines the optimal number of PCB layers based on the design requirements. Another example for mechanical design is a solution that uses machine-learning to recognize components in a design.

The learning process uses millions of different images to teach a neural network to recognize similar parts based on an example. Using camera images, the system identifies all similar parts to the example and displays the information to the engineer in real-time.

Both the learning and the inference processes use NVIDIA GPUs to accelerate this artificial intelligence system. Fujitsu’s technology demonstration should find its way into the Fujitsu Engineering Cloud services during 2017.

Autonomous Drone Survey to 3D Terrain

Yet another demanding graphics workstation application is turning image data into 3D geometry for GIS software. GeoMon delivers this service for customers using an autonomous drone, 3D photogrammetric software, and NVIDIA Quadro-based CELSIUS workstations. 

GeoMon customers working in construction and mining benefit through time and cost savings. The data capture using the drone is fast – even for larger surface areas. The processing for a typical project, which could have approximately 2500 images, can be completed within 24 hours using a single GPU. That time can be halved using a second GPU and scales almost linearly with the addition of a 3rd GPU as well. Larger projects could have 10, 000 images or more and would take over a week to process without GPU acceleration. Adding 3 powerful NVIDIA Quadro graphics cards, reduces that time by a factor of 7 to 8 times according to GeoMon.  GeoMon CTO, Sascha Heising, describes the integrated solution of drone, software and workstation. The company typically recommends the most powerful CELSIUS workstations, the CELSIUS R940 with one, two, or three NVIDIA Quadro GPUs such as the new NVIDIA Quadro P6000.

 

The PW Perspective

Virtual reality, artificial intelligence, and GPU-accelerated computing is expanding the market demand for graphics workstations. The Fujitsu CELSIUS family of workstations delivers a high-performance platform for workstation customers with support for multiple GPUs as well as other best-in-class technology. Each of these solutions was on-display for 2 days at the Fujitsu Forum 2016 in Munich. 

Go to top