In previous blog posts, we’ve covered digital twins in the automotive industry and how they can help further the development of autonomous vehicles, among other use cases. We’ve also explored how to speed up development in distributed engineering teams cost effectively with virtual desktop infrastructure (VDI) with virtual graphics processing units (vGPUs), like those from NVIDIA.

3D models are a key component for building digital twins. You might be convinced that virtualisation is great and that vGPUs are your best option for handling graphically intensive operations like 3D models, but first, what are vGPUs exactly? And how can they help you build and simulate your 3D projects?

Thanks to high performance computing (HPC) and vGPU, users, designers and developers get access to much more performance than what they would have with a regular desktop or workstation. vGPU makes it possible to run high-end graphics applications on remote devices from anywhere in the world.

What is GPU virtualisation?

When a user works on video or 3D-intensive processes such as encoding or rendering tasks, CPU-based virtual machines don’t deliver the optimal level of user experience and performance that one would expect. This is due to the way CPUs work: they are more suitable for serial instruction processing and less for parallel instruction processing, which is required for graphical operations.

Virtual GPUs allow virtual machines (VMs) to benefit from a segment of a large GPU, with performance close to GPU passthrough. This enables the sharing of very expensive high-end GPUs with minimal overhead throughout multiple VMs.

While it is possible today to use GPUs with a VM, it requires GPU passthrough that consumes the whole GPU and forces the administrator to segment it in more complex ways. For example, it requires the application to be started in a way that OpenGL calls are passed to VirtualGL. So effectively there is no real virtualisation or isolation, but more of a traditional multi-user environment.

Benefits of using vGPUs

Setting up and managing such a system makes the cost/benefit ratio less interesting compared to using vGPUs, which simplify the whole process while providing better abstraction and isolation.

The GPU passthrough splitting capability offered by vGPUs increases efficiency and reduces datacenter operating costs, as it enables optimised resource consumption. The processing power of the GPU can be used to do the computation that would previously be done by the VM’s CPU. On top of a better user experience, vGPUs allow for better performance on multiple graphical workloads (typically the case for complex image rendering).

These performance-intensive processes can run remotely on any device, including low-cost laptops, thanks to the introduction of VDI. VDI allows teams to access virtual workplaces from anywhere in the world and from any device. This also enables companies to keep sensitive data secure in highly protected centres.

In the realm of vGPUs, few names are as well-known as NVIDIA’s. From a company perspective, VDIs accelerated by NVIDIA vGPU technology provides a lot of advantages:

• Improved resource allocation.
• Enhanced security.
• No need for high-performance workstations that become obsolete after a couple of months.
• Better productivity with much higher performance, better user experience and the possibility to work from anywhere.
• Data recovery possibilities. In case of technical issues or human mistakes, it’s easy to fall back and restore a stable version.

Virtualisation and 3D CAD/CAE applications

vGPU solutions can be a great asset when it comes to graphics-intensive applications. While it’s not mandatory to build and use a highly detailed 3D model in order to monitor a specific component, computer-aided design (CAD) tools are often used during the design part of the digital twin. Having a visual model of a complete system helps you identify and analyse the impact of complex system-wide interactions.

Running the same CAD or computer-aided engineering (CAE) computations on a regular workstation would take hours just for simulating environmental conditions. Add safety scenarios and energy consumption simulations, and it would take days. You see where I’m going with this.

With a high-speed efficient network, it’s possible to run these complex algorithms on HPC clusters with very low latency and high performance. Moreover, the vGPUs can be shared across multiple VMs, allowing multiple VMs to share a GPU, splitting the overall performance of a single GPU between multiple VMs.

Automotive leaders like Honda have deployed GPU-accelerated VDI  to enhance productivity and efficiency in their production centres with NVIDIA. With graphics acceleration in the data centre, NVIDIA virtual GPUs empower teams to use CAD/CAE applications on any device — even low-cost computers.

3D models with NVIDIA vGPU and Ubuntu KVM

As software becomes essential in automotive, the competition has never been so fierce. Moreover, OEMs need to keep improving their efficiency and quality while guaranteeing safety compliance. On top of that, CAD and CAE applications are becoming more advanced and require higher computing power to work efficiently.

Virtualised environments are a big part of the automotive industry’s ongoing transformation. They are transforming the way the automotive industry is working, from the design of vehicles to their optimisation and predictive maintenance. 3D applications help build these virtual environments, but they require large-scale simulations that rely on physically accurate models. 

NVIDIA vGPU technology provides the large amounts of computing power you need to build these models effectively. Canonical, in turn, provides the best platform, enabling you to create the infrastructure that fits your needs. Ubuntu KVM, a leading hypervisor, now comes with native NVIDIA vGPU Software availability.

While NVIDIA provides industry-leading vGPU technology, Canonical helps you unlock infrastructure scalability. With Ubuntu Server, Canonical provides long-term support and maintenance. That’s why 65% of public cloud workloads use it.

Contact us to find out how you can advance your 3D modelling projects with Ubuntu and NVIDIA. 

Curious about automotive at Canonical? Check out our webpage.

Want to learn more about software-defined Vehicles? Download our guide. 

Curious about how Charmed OpenStack and NVIDIA vGPU Software work together? Watch our dedicated webinar.